yerpal sequences; a generative approach · agreement with the modal44, ... clitics, and, thus, no...
TRANSCRIPT
Mireia LLINÀS I GRAU
YERPAl SEQUENCES; A GENERATIVE APPROACH
Tesi Doctoral dirigida pel Dr. Josep Maria Brucart
( Ponent: Dr. Joaquín Domínguez )
Departament de Filolgia Anglesa i Germanistica
Facultat de Lletres
Universitat Autònoma de Barcelona
Any 1990
2.3-2 M-A sequences
2.3.2.O Introduction
I will consider M-A sequences to be those basically including a modal
or aspectual verb and an infinitive. The syntax of these sequences in
Romance has been the issue of much debate essentially because the two
adjacent verbs display a syntactic behaviour which indicates that the
structure in which they occur is "anomalous" - as will become clear in what
follows, French differs from other Romance languages in important
respects; i.e. in not allowing clitic climbing in these constructions - ; it
differs from the structure in which non-modal/aspectual verbs plus and
infinitive occur. A crucial characteristic of M-A sequences is that the
subject of the infinitive - obviously, if it is granted a clausal status - must
coincide with the subject of the M-A verb - a clear difference with the C-
sequence -:
(239)a. £tsyupiespoden JJençar eJs diners
(2 40)a. *E11yupiespoden Jes seves dones JJençsr eJs diners
b. les dones deis yupies poden JJençar eJs diners
The structure is one of Control, and thus, the possibility of analyzing
it as such is reasonable - cf. Section 2.2 - . The structure, though, displays
characteristics that indicate that it is not a normal Control structure. The
difference lies in the processes which are permitted to elements occuring in
the same clause- as will be shown in the sections to follow-: clitics may
climb to the modal/aspectual43, in impersonal si/se constructions in Italian
and -in some dialects of - Spanish objects may be preposed triggering
agreement with the modal44, whenever aspect may be expressed by
different auxiliaries there is a change of auxiliary - in Italian -, etc. This
curious syntactic behaviour was analyzed by Rizzi (19Ô2a) as being the
result of a specific rule of r$stnict.iirw£ that did away with a part of the
structure which would otherwise make the processes impossible - as in
other Control sequences -. As observed by all the proposals under review,
Rizzi's account is not tenable in the present model, although his insights
remain and have been reformulated into mechanisms which satisfy the
principles assumed in the present model. It is of uttermost importance that
none of the tests indicative of restructuring are found in English: it has no
clitics, and, thus, no impersonal constructions with subject clitics - sf/s& -
either , therefore no object preposing constructions, and aspect is only
expressed by Itave, so there is no a&£r&-jess&r$ equivalent. Restructuring
is, therefore, not a process to be considered in English.
The verbs that belong to the modal class in both English and Romance
are restricted to a small set. The sets are different and only clear-cut in
English; the set of modals in Romance is not definite and even considered
non-existent as a different category by some linguists : "... a les llengües
romàniques, és impossible delimitar formalment una categoria de verbs
modals." ( Ferrater (19Ô1) p. 11). Statements of this sort are based on the
fact that the characteristic morphological properties of English modals - and
some of their syntactic characterists - do not apply to verbs that have a
similar or equivalent meaning in Romance. For English most authors
coincide in considering (241) as an accurate list; in Catalan, (242) is only
approximative:
(241) Will/would, shall/should, can/could, may/might, must,
ought to; dare, need; is to; had better; used to
(242) poder, deure, haver de; voler, gosar
Modals in English are assumed to have the properties of other
auxiliary verbs - ci. section 2.1 - , plus other characteristic ones:
morphologically, they do not take the -s 3psg , nor have infinitive nor
gerund forms. Syntactically, they cannot co-occur, nor be preceded by other
auxiliaries.45
With respect to the morphological characteristics, Romance modals
differ: they do show inflection for person and number, as well as infinitival
and gerund morphology:
(2 43)a. * Thepresident cans change the constitution
b. Tu potscanviarla constitució
(2 44)a. * To can dance flamenco JsJn fashion
b. Poder ballar flamenco està de moda
(2 45)a. * Canning to speaA- English you can get a job
b. Podent, parlar l'anglès pots aconseguir un lloc de treball
The syntactic properties of modal verbs are not shared by modal
verbs in Catalan either - nor in Romance in general - :they can co-occur or
be adjacent (246)a., and they may be preceded by auxiliaries C247)a.:
(246)a. La Maria Jïosa deu poder deixaria feina
b. *&farla £osa may can leave her fob
(247)a. la Isabel ha pogut anar a Anglaterra
b. * lsabel has canned go to England
Nevertheless, certain relevant observations in terms of these
syntactic characteristics are made in section 2.3.2.3 - cf. Picallo (1965) and
(1990) -, where it is shown that the clearcut distinction between English
and Catalan modals is not precise in that not all types of modals display the
differences. The crucial factor involves the distinction epistemic/root
reading.
The distinction root / epistemic in the use of modals is found both in
English and Romance.^ An epistemic reading implies judging a proposition
possible or necessary; a root reading implies either ability, volition,
permission, or obligation of the subject. The following exemplify the
difference:
(1) EPISTEMIC:
a. The end of the world may i>e approaching
The end of the world must be approaching
b. la fi del mon espof. estar apropant
La fi del mon es deu estar apropant
(2)ROOr*7
a. The governments must put an end to injustice
The governments can put an end to in/ustice
b. Ets governs han de posar fi a la injusticia
Eisgovernspoden posarfi a la m/usticia
It is obvious by the use of the same lexical item in the second
example of (Da. and the first example in (2)a for English , and the first
example in ( l)b. and the second example in (2)b. for Catalan, that the same
lexical items may have different readings. This is not true for all modals as
instantiated by the verb deure in Catalan which has only an epistemic
reading (l)bv as opposed to the Spanish "equivalent" deberlo-.
(24ô)a. m fin delmundo debe (de)lestar cerca (EPISTEMIC)
b. los gobiernos dejben poner fin a Ja injusticia (ROOT)
- See section 2.3-2.3 for relevant conclusions regarding this fact-.
With regards to aspectual verbs in Romance and English, they may
add a certain temporal aspect to the complex predicate they are part of -i.e.
when they are followed by an infinitival verb^9. For this reason, they are
often considered modifiers of the predicate phrase, not imposing
restrictions on the subject of the clause . They have been analyzed as
having an adverbial function - cf. Picallo (19Ô5), Zubizarreta (19Ô2) -. Note
the possible paraphrases of the following examples:
(249)a. Es militars nan tomata amenaçar Ja democracia
b. Es militars nan amenaçat la democràcia una altra vegada
(250)a. Es politics solen dir mitges veritats
b. Es polítics molt, sovint diuen mitges veritats
(2 51 )a. Tne armvAeeps threatening democracy
b. Tne army threatened democracy once again
Aspectual verbs are usually kept apart from aspectual
auxiliaries - A? Ft?. l>e cf. Section 2.2.2 -. In this sense it is worthwhile
noting that they have non-aspectual interpretation, as in the following
examples, where thay may be interpreted as motion verbs 50,51.-
(252)a. Ha tomata dir-lique no podia viure sense ella
b. Ha vuelto a decirle que no puede vivir sin ella
(253)a. Ha vingut a dir-me que no considerava l'article prou Joo per la
seva revista
b. Ha venido a decirme que no considera mi articulo io
suficientemente Mtenopara su revista
The semantics of English aspectual verbs is, thus, equivalent to
Romance aspectual verbs, but there are, once again, no tests that would
point to a possible application of restructuring.
Note that the use of an aspectual verb does not change the relations
between the predicate and its internal/external arguments:
(254)a. Israei MisPaiestinians
b. Israeinasnot ceased toMf Palestinians
c. Israel keeps ¿iiiing Palestinians
(255)a. E/spaïsosrics s'oMden dei Tercer Mon
b. Eis països rics tornen a oMdar-se del Tercer Mon
c. Eis països rics seien oMdar~se dei Tercer Mon
As is suggested by the preceding observations, the kind of syntactic
variation in M-A sequences is more fundamental than the contrasts
observed in C- sequences - cf. 2.3.1 - : it not only involves a different order
of constituents but there may be reasons -cf. especially 2.3.2.3 - to posit a
distinct category for some modals - a "traditional" consideration for English
modals. As already observed in section 2.2.2, modals are assumed to be
generated in INFL - cf. especially Chomsky ( 19Ô1) p. 140 fn. 26* -. Therefore
sequences of modals and infinitives in English are not considered in this
brief introduction to the proposals considered in this section, as they do not
constitute complex predicates.
The relevant factor oí M-A sequences - as with C-sequences - is that
they consitute an instance oí V+V:
(256)a. Lady Macbeth va voter convencer ai seu marit.
b. Lady Macbeth qwso convencer a su marido
c. Lady Macbeth¿a voliitio convincere ii suo sposo
in the sections that loiiow, the authors focus on the explanation oí
the processes - if any - that relate the two verbs and allow them to
function as a unit with respect to certain syntactic phenomena -
alternatively, the process erases the clausal boundary between the two
verbs -. The validity of the processes will ultimately depend on whether
they satisfy the principles in the grammar, and whether they are,
independently required, since the structure gives way to no changes - cf.
C-sequences and the non-standard word order -. In (256) the postulation of
a process is PF vacuous, in (2575, cliticization of the embedded object is
seen as a consequence and, therefore a signal, of the application of the
process - cf. Section 2.3.2.1, 2.3-2.2 -:
(257)a. Lady Macbeth eJ va voter convencer
b. Lady Macbeth Jo quiso convencer
c. Lady Macbeth J' ha voiutto convincere
Note that French does not allow the same structure. A fact which is
not explained in the proposals in this section. Several of the proposals
summarized attempt explanations to this fact -especially section 2.4 -:
(256)a. Lady Macbeth a voulu convaincre son mari
b. Lady Macbeth a voulu Je convaincre
c. * Lady Mscbotn J 'avoulu convaincre
The proposals reviewed are representative of different alternative
analyses of M-A sequences. The authors who postulate a biclausal status for
the structures - Rizzi (1902a) and Burzio (19Ô6) - sustain arguments for a
clausal subcategorization and subsequent processes of deletion - Rizzi - or
movement - Burzio -. Deletion is a much too powerful mechanism , barred
in the present framework as pointed out in several of the following
sections.
Those authors defending a monosentential analysis need not recur to
movement or deletion but must account for the non-existence of a subject
position in the infinitive complement. 52
2.3-2-1 A Restructuring Rule ( Rizzi 1902a)
In his well-known article "A restructuring riûe" Rizzi examines the
status and structure of certain verbs taking infinitival complements , and
puts forward a rule -restructuring - which these verbs optionally trigger.
His analysis is set within a pre-GB model; i.e. he makes use of some
mechanisms which have since been abandoned or reformulated. One of
these is the Specified Subject Condition (SSC) - cf. 2.3.1 -. As observed in
other sections, within the new framework, the effects of such a condition
follow from other principles of the grammar - for instance, binding . In the
model in which Rizzi develops his proposal the transformational component
had not yet been reduced to its limit so he makes use of specific
transformational rules for specific structures. One of such rules, which has
been explained above, is Clitic Placement - cf. Kayne(1975). In spite of this
background, Rizzi does not actually give a complete characterization -
Structurai Description and Structurai Change - in transformational terms
of the rule that he posits -moreover, he gives the final format in a footnote
( cf. p.47, fn. 42 ). The fact that the transformational component contained
a series of different rules favoured the reference to extrinsic ordering to
account for phenomena which could not be explained otherwise - Rizzi,
though, argues against conceivable alternative explanations of his data
through the use of this strategy. It is obvious that in the present model
Rizzi "s proposal needs reformulation, which is the case of some of the
analyses sketched in the following sections.
Rizzi" s main goal is to characterise a set of main verbs in Italian
which show a distinctive behaviour with respect to certain - previously
unrelated - syntactic processes. The set of verbs includes modais.
sspectuais and motion verbs. The syntactic processes the explanation of
which he unifies are basically cliticization, object preposing in impersonal
"si" sentences, optionality in auxiliary selection. Considering the structure
which makes these syntactic processes possible, he postulates a
derivational account of the data; i.e. the output structure is different from
the structure at the outset of the derivation. This last remark is essential in
his explanation and in order to prove that a specific structure is involved,
he tries these processes by making use of constituency tests; certain
syntactic operations only apply to whole contituents so the structure
resulting from the application of restructuring will either prevent or allow
these processes to operate. The processes involved are: wn-movement {Piecf
Piping), cleft sentence formation. Pignt Node Raising (RNP¿ zxiâCompJex
NP-Süift Briefly, if restructuring applies, elements which were part of
different constituents - two verbs - become united into one constituent, "a
verbal complex"; and elements which were in the same constituent - an
embedded clause - become members of different constituents. The
processes mentioned are correctly predicted to apply only when
restructuring has not applied. The only way to verify the fact that
restructuring has applied is by alluding to the set of syntactic processes
that are particular to verbs which trigger this rule, thus, "constituency
tests" will only be possible if clitics have not moved, objects in impersonal
"si" constructions have not preposed, and there has been no auxiliary
change in a structure which would otherwise allow it. These predictions are
briefly reviewed below.
Restructuring, as Rizzi puts it: "optionally transformis] an underlying
bisentential structure into a simple sentence, creating a unique verbal
complex consisting of the main verb and the embedded verb." (p.2) In
other words, sequences such as the ones in (259) have two possible
structures: a bisentential one - if the rule does not apply -, which implies
the presence of a clausal node intervening between the two verbs; and a
monosentential structure - ií the rule applies -, which results in the loss of
the clausal node between the two verbs:
(259) a. voter faro/poter fare/dover fare
b. continuare a faro / fJnJscere di fare / starperfare
c. venire a faro /andaré a faro
The following structure is given by Rizzi to illustrate the process:
(260)
ojanm fvi1eve presentare! Ja a Francesco
(16)
The first verb is a modal and for this reason, it triggers the
application of the rule and the two verbs are reanalyzed as a verbal
complex with no clausal boundary nor subject position intervening.53
The first of the syntactic processes that Rizzi deems possible because
of the application of restructuring is cliticization. In (260) the fact that the
two verbs are reanalyzed as one verbal complex allows the clitic to take a
"long step" and cliticize to the first verb of the complex instead of cliticizing
to the infinitive:
(2 61 ) a. Gianni Is dove presentare a Francesco (17)
b. Gianni dove presentarla_a Francesco ( 1 ô)
The rule involved is Clitic Placement (CP), which has the following
format ( p.3 in Rizzi( 1902a)) - cf also 2.1 -:
(262) vbl - V - vbl - PRO - vbl
1 2 3 4 • 5 - ^
1 4+2 3 / 5
The crucial condition concerning Rizzi's proposal is that terms 2 and 4
must be clausemates; the PRO - here: clitic pronoun- and the verb to which
it clicizes cannot be separated by a clausal boundary. Therefore, if (259)
did not imply an erasure of the clausal boundary dominating the second
verb, CP would be predicted impossible, contrary to fact. Within the model
then used, such a condition followed from the SSC: a specified subject is
present in the embedded clause (PRO) which prevents the application of
any rule. If CP applies, the application of restructuring is indispensable and
compulsory. Nevertheless, if CP does not apply, the application of
restructuring is, in principle, optional since it applies vacuously.54
Nevertheless, Rizzi assumes that restructuring does not apply if the clitic
does not take the "long step". This is precisely what his "constituency tests"
suggest, as will be illustrated below.
Another syntactic process indicative of the application of
restructuring is object preposing in impersonal "si" sentences - cf. also
section 2.1 - . I will not review the analyses of this construction but simply
point out the aspects relevant to Rizzi"s proposal. A sentence like (263)
contains a subject clitic "si", which does not - at least at this stage of the
derivation - occupy the subject position but has cliticized to the verb
adjacent to it.
(263) Si dorme troppopoco (50)
if the verb is transitive, Italian has a special construction which
allows the preoposing of the object to preverbal position:
(264) a. Si contruisce treppe casein quests città
b. Troppe case si contriuscono in quests città (57)
As (264)b. illustrates, this movement triggers agreement with the
verb; the preposed object functions as the subject. If there are two verbs in
a sentence, this transformation is only possible with the set of verbs
relevant for restructuring:
(2 65) a. *Le nuove casepopoiari si sono promesse di construire
entro un anno (59b.)
b. Jprobiemi principan si continuano a dimenticare (63b.)
The explanation for this is also given by the fact that restructuring
allows an apparent violation of the SSC by the rule of object preposing; the
clausal boundary - and, thus, the subject - is erased only if the matrix verb
is a member of the class of triggering verbs.55
The last basic piece of evidence which Rizzi brings to bear upon the
application of restructuring is a particular behaviour of the class of
"triggering" verbs with respect to the choice of auxiliary . In Italian, there
are two auxiliary verbs required by different lexical verbs: essere.. and
avere:
(266)a. fiero J¡a/*& volute cuesto libro (71)
b. Piero *hs/è venuto con noi (72a.)
As (266)a. illustrates, modals functioning as main verbs take avere.
Nevertheless, in constructions where modals take an infinitival complement
which requires essore, they no longer impose this requirement:
(267)a. Pioro lia/e voluto/potiito/dovuto venire con noi (72b.)
b. Piero M^—À. promesse di venire con noi (77)
Such examples show that a process of auxiliary change (avere—
essere) is only possible with verbs which allow restructuring. It illustrates
the fact that modals in a structure like (267) - as opposed to (266)a. -
function as auxiliary verbs ; they are part of a verbal complex in which it is
the embedded verb that imposes its lexical requirements.
Another important issue in Rizzi "s proposal are his "constituency
tests" which, as mentioned above, are referred to as evidence for the fact
that restructuring turns constituent-mates into members of different
constituents. Namely, that presentaré a Francesco in (2 6 Da. is not a
constituent but that presentarla a Francesco in (26l)b. is a constituent;
that dimendicare i problem! in (265)b. is not a constituent, but that the
same structure without agreement -si continua a [dimendicare i problem!I-
is a constituent; and that venire con noi in (267) a. is a constituent only if
the ayer* occurs as an auxiliary. What is crucially at stake is that the
application of restructuring precludes any rule whose application hinges on
the fact that the infinitival complement is a clause; it is only a clause if it
has not been restructured
The following are some examples of the different processes which
indicate that restructuring has applied, having undergone one of the
"constituency tests" mentioned; their grammatical counterparts show that if
restructuring does not apply, the structure meets the structural description
of these "constituency tests": - cf. Chomsky (1972), Ross (1967), Postal
(1974), among others, for a description of these transformations -:5&
Clitic Placement:
-Wh-movement:
(26ô)a. Questl argomentl. a parlarti del quail verre s/p/u presto....
b* Çuesùargomentl, a par Jare del quail ti verre al plu preste...
(10)
- Cleft S Formation:
(2 bQ)a.I"propplo arlpertagJJ 1 seldl die ste andando, stal tranquille/
i>.*F'propplo arlpertare 1 seldl che glî ste andando....
(30)
- Right Node Raising:
(270) a. Mario sinceramente vorrebbe - ma a mloparère nonpotrà
mal - pagargJi Meramente 11 suo debite
b* Mario sinceramente gil vorrebbe - ma a mlo parère non
potra mal -pagare Intoramente 11 suo debite
(37)
-Complex NP Shift:
(2 71 ) a. Fra quaJcne giorno, verrò a Firenze adesporti Ja mía idea
b. *Fra quaJcne giorno, ti verrò a Firenze ad esporre Ja mía idea
(41)
Before closing this brief review on Rizzi's restructuring proposal, I
will give the actual format of the rule :
(272) vbl Vx y ] (COMPL) V vbl ~>vbl Vx(COMP) V v- ] vbl
(where Vx is a V member of the "triggering" class of verbs)
(p.47fn.42)
Rizzi grants the node dominating the verbal complex a V status on
the basis of examples like (273) which, if we assumed the node had an X-0
status, would " deprive the notion of lexical category of its content" (p.3Ô),
allowing other lexical items to intrude between the parts of an X-0
category.
(273)a. Lo verre subito a scrivere
b. GJi stessd error/ si continuano stupidamente a commetere
c. Maria è dovuta immediatamente tornare a casa
(144)
Note that all of these examples are examples of restructured
structures. The basic argument to posit such a projection dominating two
verbal elements comes from the observation that it is independently
needed in Italian for sequences of auxiliary-verb, which may, furthermore,
be interrupted by other lexical items :
(274) a.j%> subite scrJíte aJfraacesw
b. Man's ê Jmm&diatamente ternata a casa
(145)
The analysis sketched above inevitably gave rise to many proposals
and counterproposals. In the following sections I will summarize some of
the most salient related analyses within more recent models than that
within which Rizzi ( 1962a) was posited. Here though I want to mention the
article Hernanz & Rigau (164) who assumed Rizzi"s analysis, applying it to
Catalan and Spanish; i.e. they noted that the same type of verbal complexes
are found in both languages - cf. 2.1.2 for a comment on the nature of the
verbs which are assumed to undergo this process -. They put forth more
arguments and show that M-A verbs differ not only from proper main
verbs but also from auxiliaries; they are "double natured". Following Rizzi,
they claim that a rule like restructuring would explain their characteristic
behaviour. Nevertheless, they note that Rizzi's rule must be reformulated in
terms of the present theory because, as it stands - as presented in this
section -, it does not conform to the principles of the theory. One of these
principles is the Projection Principle which ensures that lexical properties
of items are kept throughout the derivation - cf. section 1.2.1 -. Notice that
a rule which erases or destroys structure stands in sharp violation of such a
principle. Much of the subsequent work on "restructuring" verbs attempted
to avoid this.
2.3-2.2 Restructuring as VP-movement (Burzio 1966)
Burzio "s analysis of restructuring constructions shares many of the
insights of Rizzi (19ô2a). He considers that restructuring constructions are
instances of "complex predicates" as evidenced by the constructions that
Rizzi accounts for; namely, Clitic Climbing (CL CL), Long Object preposing
(Long OP.) in sz/S*? impersonal constructions, and Change of Auxiliary (CA)
- avère— ess&re where both auxiliaries may realize aspect - . As shown in
the preceding section, these phenomena occur with the same class of verbs
- "restructuring" verbs -; they share structural characteristics - cf. the
constituency tests in Rizzi (19ô2a) -; and they interact - i.e. if one occurs,
the others do, too. What these constructions are a reflex of is the fact that
the resulting structure is a "complex predicate": there is " a closer than
usual relation between the main V and the infinitive" (p.32Ô). In the
framework in which Burzio makes his proposals, the "structural anomaly"
of these constructions would imply an exceptional non-violation of the
Binding Principle A - in the previous section it was shown that in Rizzi "s
framework, the relevant principle is the SSC.
In line with Rizzi's proposal, the special cohesion of the verbs is
assumed to be the result of a syntactic derivation, as opposed a base-
generated complex predicate; nevertheless, the type of derivation is not a
reanalysis of the structure - with the consequent subject deletion -, but
rather movement of the embedded clause VP. Burzio also diverges from
Rizzi in the assumption that causative sequences and restructuring
sequences may be accounted for in a parallel fashion and that the
difference follows from the different structures to which the rule of VP-
movement applies.
The similarity of restructured and non-restructured structures with
respect to selectional restrictions is a crucial fact for positing a syntactic
derivation versus base-generation. Within Burzio's classification of verbs,
those which undergo restructuring belong to three different types:
(275) ERGATIVE: andaré, venire
RAISING: dovere, potere, cominciare, continuare, sater (per),
sembrare
CONTROL: volere, sapere, cominciare, continuare (ô)
The following show the three types of input and output structures to
restructuring as VP -movement:
(276)a. Giovannij va t][PRO] a prendere]]iibro/
b. Giovanni] va [a prenderen iibro It] [PRO] —/
(277)a. Giovannij dovrebbe [ i] prendere ii iibro I
b. Giovannis dovrebbe [prendere ii iibro! [ t] — /
(27ô)a. Giovanni] vorebbe [PRO] prendere ü iibro I
b. Giovanni] vorrebbe [prendere ii iibro ¡[PRO] —/ (9)
Note that they crucially involve configurations with coindexed
subjects. Many differences with causative constructions will follow from
this; i.e. the fact that the two subjects are coindexed only in restructured
constructions - cf. (296) -(296).
In a structure like (277), the matrix subject is subject to the
selectional restrictions of the embedded verb, as clearly illustrated by
(279) . This is a typical argument for Raising, and it differs for Control
structures, where there is a "double" dependence - i.e. the matrix subject
must obey selectional restrictions of both, the embedded verb and the main
verb, as (2Ô0), (2Ô1) and (262) illustrate:
(279) a. IIlibro dovrobbo essore portato da Giovanni
b. L 'acqua dovrebbe scorrere
c. Dovrebbe plovere
d. Dovrebbe risultare che Giovanni non c'era ( 1 ô)
(2ô0)a. * Il libro viene ad essore pórtate da Giovanni
b. *L acqua viene a scorrere
c. * Viene apiovere
d. * Viene a risultare che Giovanni non c'era ( 19)
(2 Ô1 )a. *lllivro vuole essorepórtate da Giovanni
b. *L 'acqua vuole scorrere
c. * Vuole piovere
d. * Vuolerisultare che Giovaninon c'era (20)
(2 62 )a. * Giovanni viene ad essore lotto da Mario
b. * Giovanni vuole piovere (21)
The relevant data for identifying restructured and non-restructured
complex predicates is exemplified by the following sentences, in which clitic
climbingi CL CD indicates that restructuring has applied, and selectional
restrictions apply in parallel to non-restructured complexes:
(2ô3)a. Il libro glidovrebbe essore portato
b. L 'acqua vi dovrebbe scorrere
c. Cf dovrebbe piovere dentro (22)
200
Burzio invalidates three possible alternatives to a syntactic
derivation of these complex predicates: that restructuring verbs are like
English modals; that restructuring constructions involve base-generated
VP-complements; and that restructuring complexes are base-generated
complex verbs. He notes that although these would all account for CL CL
and Long OP. correctly - i.e. Binding Principle A would not block any of
these processes-, they would fail to account for the similarity in terms of
selectional restrictions; i.e. the main verb failing to impose restrictions
would not be explained if it were taken as the head of the verbal complex,
and thus a primary predicate - as in (2Ô3).
Burzio "s specific proposal of VP-movement is founded on, basically,
evidence from the restructuring ergative verbs anâare and vemre . He
claims that the effect of restructuring on these verbs provides sufficient
motivation for a VP-movement formulation which changes the linear order
without implying an elimination of the embedded subject position. As has
already been mentioned - cf. the introduction to this section - the deletion
of the subject position is not in line with the present framework. Burzio
notes two basic facts that point towards the existence of a subject position
in the restructured complexes: one is their interpretation - the selectional
restriction observations made above -, and the other one is the fact that the
existence of a trace in subject position - which is not properly bound, and
thus, violates the ECP - accounts for the ungrammatically of (2Ô4) ,
otherwise unexplained. Note that votere is nota Raising verb and, thus, the
trace cannot be properly governed even if there is no restructuring, but if
restructuring deleted the trace, the ungrammatically would not be
predicted.
(2 64) *Inostn atfetij sf vorrebtero [ vlnœrelitj — / (43)
The interaction of restructuring ergative verbs and a fare -
construction involving a VP-complement with CL CL of the subject of the
ergative gives evidence for the fact that the object position - the original
position of the subject of the ergative - has not been deleted, but rather
that the VP of the embedded clause has been moved. The dativi2ation
process, mentioned in Section 2.3-1-2 in relation to the causative
constructions, applies in this context: the NP II libro becomes an object of
the ergative verb, and of the causative verb as well - all the boundaries
that separate fare and the NP are VPs . The ec of the clitic becomes a
second object of fare and the context for dativization arises, accounting for
the occurrence of git and not lo:
(2Ô5) GUj faro fyp andaré [yp a prendere il libro / ft e /
ÍSPWj — / (36)
The following change in structure is what accounts for dativization;
NP2 induces dativization of NP i:
(266)a. IvPlVi NPiIsPROÍVP2V2 NP2 111
t>. IVPl Vl [ V2 NP2 1 NPl Is PRO ---]]] (37)
The lack of S SC effects in sentences like the following is another
argument for VP-movement; it follows from the movement of VP out of the
clause, out of the domain of the "specified subject":
(2Ô7)a. Marioloj vuoleleggere fiel (la.)
b. Questilibrij si volevanopropio leggere tj (2a.)
The fact that the subject of clauses embedded under restructuring
verbs is always null is assumed to be a consequence of the fact that there is
no way for it to acquire Case: verbs like voters are not S'-deletion verbs, so
there is no government ,(2ôÔ)a.; and verbs like potere. sembrare trigger S'-
deletion but do not assign Case, (2ôÔ)b.:
(2ÔÔ) a. * Maris vuoie fme participare I
b. *ie/puo/sembra i me partecipare / (47)
This observation carries over to restructured constructions where the
main verb still fails to govern and/or assign Case to the embedded subject:
(2Ô9) a. * Maria vuoie [yp partecipare /is me — /
b. * ie/ puó/sembra fyp partecipare/is me — / (46)
The lack of dativization of the subject of the embedded verb - as
opposed to causative structures - is accounted by the failure of application
of the dativization process. This process only applies if both objects are
governed by the main verb or assigned Case by it; the non-S'-deletion
accounts for (290)a. and the lack of Case assignment accounts for (290)b:
(290)a. * Maria vuoie [yp ieggere ii iibro /is a Giovanni — /
b. * i e/ puó/sembra fyp ieggere ii iibro / is a Giovanni — /
The lack of phonetically realized subjects follows from Case
requirements, and the assumption of the presence of a syntactic subject at
all levels is not undermined by this fact. Therefore, restructured and non-
restructured structures are alike in D-structure; there is no deletion of the
subject position. This also follows form the Projection Principle and the fact
tnat semantic interpretation is derived from S-structure. The existence of
the subject at all levels, moreover, is necessary to ensure the correct
distribution of Raising and Control.
Some similarities and some differences between
restructuring and causatives
In Burzio's analysis, the similarities observed between restructuring
complexes and causative constructions support the VP-movement analysis
in that they are both derived structures; the differences also corroborate it
indirectly in that they can be attributed to the distinct properties of each of
the D-structures to which VP-movement applies; namely, coreference
between main and embedded subject in restructuring.
Two of the similarities that Burzio observes are the distribution CL CL
and past participle agreement. Clitics in this construction are assumed to be
base-generated in the embedded verb and moved to the main verb after or
in conjunction with VP-movementBinding Principle A is, thus, not violated:
(2 9 Da. Lij ho fatii [yp JoggereQej] fs a Mario — /
b. Lij ho voiuti ÍVP teggere fi # JffPRO — / (56)
(291) also exemplifies past participle agreement, which takes place
between the past participle causative or restructuring verb and an
argument of the embedded verb. These arguments have, thus, been
reanalyzed as dependents of the matrix verb as well.
Clitic climbing exemplifies one of the two possible relations that may
hold between an element in the matrix clause and an element in the
embedded clause - cf. (293). Past participle agreement is one of the two
phenomena that are triggered when this relation holds - cf. (292); the other
204
phenomenon is ¿äaare - assignment (E-assignment). These two phenomena
directly bear upon one of the differences between causatives and
restructuring constructions: the fact that restructuring with ergative verbs
is possible - recall that VP-movement analysis of causatives does not allow
ergative causative constructions because an NP-trace relation must be met
at all levels ; these are analyzed as FP constructions -.
Past participle agreement and E-assignment are defined by Burzio as
in (292); (293) are the possible relations and the rule that they trigger:
(292) a. Auxiliary E is assigned when there is a relation of a certain
type between the subject and either a clitic or a direct object
b. A pp will agree with an element holding a relation of a
certain type with its direct object
(69)
(293) a. NP cl V . . . E only
b. . . . cl V NP . . . pp agreement only
c. NP V NP . . . Both E and pp agreement (70)
In restructuring constructions, pp agreement is triggered as in (291)
above and (294), which shows that the embedded object does not cease to
be an object of the embedded verb even if it also becomes an object of the
main verb:
(2 94) Lij vorreiÍVP avergia tetti fj<? / / is PRO — / (74)
*
The fact that restructuring verbs can occur with ergative verbs is due
to the fact that the trace of the object of the ergative is properly bound at
S-structure, after VP-movement, as a result of the basic coindexation - i.e.
tiie Control relation - of the matrix subject and the embedded subject; the
trace in object position is also coindexed with the matrix subject and, thus,
properly bound - i.e. it has a c-commanding antecedent at S-structure -:
(295) a. fj&fsij sarebbe volute { PKOj andaré tj/
b. fje/sij sarebbe voiuti {yp andaré tj/fPffij — / (76)
This process is what Burzio calls Subject Substitution, which is only
possible when the two subjects are coindexed. The distribution of E -
assignment and pp agreement is predicted by this process: this is one of the
possible relations which will trigger these phenomena-(293)c. E-assignment
applies when there is restructuring - a fact observed by Rizzi, but explained
by Burzio:
(296) a. Nolj avremmo voiutols PROj andaré tj /
b. Noij saremmo voluti [yp andaré tiffs PRO] — / (60)
And the difference with the causative construction follows in that in
causatives, there is no relation which can trigger pp agreement (297) nor E
-assignment (2 95):
(297) fjefsij sarebbe fatto/ *fatti [yp andaré Giovanni I (79)
(2 9ò)J]/oj avremmo fatto/ *saremmo fatti [yp andaré Giovanni / (Ô1 )
The differences are thus predicted: in causative constructions only
certain complements may be embedded under fare ( Fl) - i.e. those not
involving an NP-trace relationdhip - (299)a vs. (299)b. -; in restructuring
constructions there is no bifurcation since they all involve a relationship
between the two subjects - either Raising or Control - and, thus, allow
Subject Substitution, (300). Recall that (200)a. is possible because
reconstruction licenses the relation between the lexical anaphor and its
antecedent - cf. section 2.3-1-2:
(299) a. Giovanni ne/ farà {yp invitare una [j e I dascunoj I
ÍS ai suoi amidj — /
b* Maria gJij farà [yp esserepreséntate tj fief/
ÍS (à) Giovanni/ — / (95)
(300)a. / suoi amid se vorebbero /potrebbero /andrebbero ad
invitare una dascuno
b. Giovannigli vorrebbe /potrebbe /andrebbe ad
essere presenta to
The different properties of each of the two constructions, thus, follow
from their different initial structures, but there is a common formulation
that can be assumed for both, a fact not considered by Rizzi. Burzio puts
forth the proposal that VP-complementation is a marked option - which he
assumes for FP construction (cf. Section 2.3-1-2 ), and that restructuring and
causative structures only exist in a language if they are "minimally
productive" ; i.e. they may occur with transitive, intransitive and ergative
verbs. Since Subject Substitution is only possible with restructuring verbs,
in order to allow ergatives, causative constructions may take VP-
compiements to satisfy the "minimal productivity" requirement. Another
difference that follows from general principles is the fact that the causative
nile is obligatory and the restructuring rule is not obligatory: only the
former is needed for Case-assignment to the embedded subject.
2.3-2-3 Monosentential analyses of restructuring complexes
Picallo (1965)
The analysis Picallo (1965) provides of modal and aspectual verbs is
related to her proposal for redefining the notion of Governing Category,
which captures the special properties of subjunctive clauses with respect to
the coreference possibilities of their null pronominal subjects; i.e. pro
cannot corefer with an element in the matrix clause. This implies that
Binding Principle B applies on the more inclusive domain; the Governing
Category of the pronominal in the subject position of the embedded clause
is not its S - cf. (301)- (303) -. Picallo redefines Governing Category in
terms of Tense chains in multisentential structures where the verbal head
of the embedded clause is "tense-dependent" - the subjunctive -. Elements
subject to the Binding Conditions must be linked to an accessible Subject
within a Tense chain. Structures containing modal verbs in subjunctive
clauses behave differently as regards opacity - cf. (304) -(306). They allow
the subject pronominal to corefer with the matrix subject. Picallo refers to
this as the "opacity-inducing" property of modals ; (epistemic) modals seem
to delimit a binding domain:
(301) *[la seva/ esperança que fproi partes amb eil //
anava disminuint
(302 ) * proi sentien que ípny produissin una faisà impressió/
(303) * Tj 'agradava que fproj ei consideressis amic /
(304) [La sevai esperança que [proi pogués parlar amb eii //
anava disminuint
(305) proj sentien que fproi deguessin produir una falsa impressió/
(306) Tj 'agradava que (proi poguessis considerar-lo amic I
(D-(6)
The explanation to this exceptional behaviour of only epistemic
modal verbs is contingent on their analysis as INFL elements. It is assumed
that no Tense chain is formed linking the modal with the tense in the
matrix clause because the modal in INFL, its head, cannot be anaphorically
related to the head of the upper clause. This behaviour is opposed to root
modals and aspectuals (30?) - (309); the coreference possibilities of their
null pronominal subjects are as in (30D-Í303):
(307) * La Isabel] esperava que {proj hi volgués anar I
(30ô) * [La seva/ Insistència que [proi la comencessin a demolir II
* era dubtosa
(309) * En Joan] desitjava que [proi tornes a venir I
(9),(6),(7)
Since Picallo's reformulation of the binding domain does not bear
directly on this thesis I will not review it further, but will proceed to
summarize her analysis of modals and aspectual verbs.
The basic insight in the proposal under review is that complexes of
restructuring verbs and infinitivals are not bisentential at any level of the
grammar; they are monosentential base-generated verbal complexes.
Nevertheless, the basic contrast in terms of epistemic vs. root
interpretations is accounted for by the possibility of modals of occurring in
different positions in phrase structure. This carries over to aspectual verbs,
which are analyzed as occupying the same structural position as root
modals verbs. The phrase-structure contains the following positions for
modals and aspectuals:
(310)5? INFL"
NP INFI/
[Tense,AGR] INFLÛ Vf
Modal
Aux Vi'
' v°j \ Modal V°i
(53)
As the subscripts and bar-projections indicate: infinitives in
restructuring complexes are the heads of the verbal projection ( Vi ) ; root
modals (Vj) are adjoined to a verbal projection; and epistemic modals
(INFL0) are the heads of INFL. Therefore, modal verbs do not head a verbal
projection.
As seen in the preceding sections, the main arguments against a
derived analysis for restructuring complexes are based on observations on
selectional restrictions. In contrast with Burzio (19Ô6) - who uses these
observations precisely to distinguish between Raising and Control - in
Picallo (19Ô5), the Raising / Control distinction is seen as irrelevant for
restructuring complexes . Firstly, the distribution of GO/HM cliticization
shows that if the structures were analyzed as instances of Control, a
violation of the Projection Principle would arise. Secondly, Rizzi (1966)
locality condition on the formation of chains shows that modals are not
/ \
relevant for the determination of the presence of a derived subject - i.e. of
a Raising predicate, as will be seen below.
Selectional restrictions show a distinction between modal verbs: voler
assigns an external theta-role, but poder does not. Voler is, thus, analyzed
as a Control verb, whereas poder is a raising predicate - cf. section 2.3.2.2,
Burzio (1956) -. (311) and (312) illustrate this. Nevertheless, poder has
another reading in which it contributes to the selectional restrictions that
must be satisfied by the nominative NP, (313) has a double interpretation.
This fact would require a double subcategorization frame for modal verbs.
Picallo, instead, argues in favour of a double-position in the phrase-
structure, and against a raising or Control analysis:
(311) * Les pedres ni* volen caure [el1 (24)
(312) Les pedres ni J poden caure Íe/J (25)
(313) Ei lladre pogué entrar per la finestra (2 Ô)
The distribution oí en/ne cliticization shows that in a restructuring
complex, the infinitive is the dominant head in terms of thematic structure;
therefore, a Control analysis of the modal V is undesirable as the " formal
manifestations of the thematic structure" of the modal verb would change -
being invalidated by the infinitive - leading to a violation of the Projection
Principle. The relevant properties of en/ne cliticization are the folowing: it
stands for the head of an N' constituent of a quantified NP which is an
internal argument - either a direct object (315) or the nominative subject
of an ergative verb (314)-:
(314)a. Han sortit algunes persones
b. Nnan sortit, algunes
c. * Han sortit algunes (29)
(315)a. Sempre compra molts llibres el Guillem
b. Sempre en compra molts el Guillem
It is obligatory if there is wh-movement of the specifier of the
internal argument:
(316) a. Quants n han sortit
b. * Quants han sortit (30)
it is disaiiowed when the quantified NP is in preverbal position
independently of the status of the NP - internal / external -:
(317)a. * Algunes n'han sortit (31)b.
b. * Alguns n'han dormit (32 )b.
It is disallowed when the postverbal quantified NP is not an internal
argument, but a postposed external argument:
(31ô)a. Parlaven molts nois
X>.* En parlaven molts (33)a.,b.
(319)a. Volien llibres alguns amics teus
b. *£n volien llibres alguns (34)a.,b
The distribution of this phenomenon depends on the external theta-
role assigning properties of the verb: it is only allowed if the postverbal NP
is not an external argument. The distribution of en/he cliticization in
restructuring verbal complexes is incorrectly predicted if the modal verbs
are assumed to be verbal heads imposing their selectional restrictions on
the arguments of the verbal complex - i.e. Control predicates -; the facts
212
bear out Picallo's analysis of modal verbs as non-heads and infinitive
"complements" as the heads of the verbal complex, en /ne cliticization is
allowed only if the infinitive is an ergative - independently of the nature of
the restructuring verb as an ergative, Raising or Control predicate -:
(32 0)a. * Quants te ¿n 'arribaren a escriure cartes fe ¡J
b. Quants tJ ' arribaren a escriure cartes fe/J (40)
(321 )a. *HJ1 tornaven a créijser moltes fe fJ
b. N'hiJ tornaven a crëïsermoltes fe fJ (41 )
(32 2 )a. *Quants te ¿ n comencen a escriiwe cartes fe fJ
b. Quants et J comencen a escrJure cartes fe/J (42)
(32 3)a. N'hi1pogueren entrar s/s fe P
b. * Hi1 pogueren entrar sis fe I1 (44)
(324)a. Quants n ni * vollen arribar [el1
b. * Quants ni i volien arribar fe f¿ (45)
Cliticization is only allowed of internal arguments, so if the only
argument of a verb is en/ne cliticized, the verb must be ergative.
Neverthelss, Control verbs like voJer require en/ne cliticization as in
(324). The contradiction is that in restructuring constructions, Control verbs
behave semantically like Control verbs , but syntactically like Raising
predicates. As was pointed out above, this is interpreted as a violation of
the Projection Principle. Picallo follows Zubizarreta (1962) in attributing to
these modal verbs the function of secondary or adjunct predicates, although
she diverges from this proposal in not granting modals a double
simultaneous structure - cf. section 2.3.1.3 for the equivalent proposal for
causatives -.
Before considering the adjunct-predicate status of root modals, I will
briefly mention how Picallo invalidates a raising analysis of some modals
by referring to Rizzi (19Ô6) locality condition on chain formation.
Configuration (325) is ungrammatical by condition (326), which disallows
interruption of a chain by coindexed elements:
(325) *NPj... dj... [e//... fe¡i (57)
(326) C = (@i, ...@n ) is a chain iff 1 i n
@ i is the local binder of @ i+n
(i) @ is a binder of ß iff, for @, ß = any category,
@ and ß are coindexed and @ c-commands ß.
(ii) @ is the local binder of ß iff @ is a binder of ß,
and there is no ¥ such that ¥ is a binder of ß, and
¥ is not a binder of @. (66)
It accounts for ungrammatical structures with anaphoric clitics and
NP movement - i.e. in configurations with derived subjects -, such as (327)
and (32 ô):
(32 7) * Eis nostres amies/ esj foren presentats fe /j fe // (56)
(32 ô) * Ei Joanj es/ sembla [ e Ij intelJlgent fefj (62)
(329) shows that analyses of modal verbs as raising predicates are
not consistent with the locality condition on chains; contrary to predictions
of these analyses, an anaphoric clitic in a restructuring construction is
allowed:
(329) En Joan es deu afaitar
(230) En Pere i en Gerard es leiten considerar inteiiigents (67)
The condition on chains is not violated if epistemic verbs are
generated in INFL - as in (310) above - and not considered raising
predicates. As with en/ne cliticization, the infinitive is shown to be the
head which determines the argument structure of the complex. Anaphoric
clitics are not allowed in structures with a modal and a non-external theta-
role assigning predicate:
(331) * En foam esj devia semblar fe ¿ intelJigent fe ¿ (71)
(332 ) *Els teus amksj sjñavien de ser presentats fe Jj fe ¿ (72 )
(333) * En Joan] esj podía semblar fe // un geni fe ¡i (73)
The parallel behaviour of aspectuals implies that these are not
raising predicates either:
(334) Enjeani esi començava a afaitar íe u
(335) Enjoani esi solia afaitar fe ¡i (74)
(336) * Els meus arnicsj esj començaven a ser presentats fe Jj fe ¿
(337) * En Joanj esj solia semblar fe ¡i inteüigent fefj (75)
On the basis of the fact that they do impose restrictions on the
nominative subject (33ôb. vs. 339b), but that they are within the VP when
they have a root interpretation ( 340a.), root modals are granted an
internal VP position as in (310) above. Note that the fact that they are
within VP is argued on the basis of the distribution of aspectual markers
with modals. Modals only have a root reading if they are preceded by
aspect markers (340a), if they are followed by aspect markers, then they
may only have an epistemic reading (340b):
(336) a. Et quadre em va impressionar
b. *E1quadre em va voler impressionar (149)
(339)a. En Joan em va impressionar
b. En Joan em va voler impressionar ( 150)
(340)a. En Joan na pogut anar al cinema
b. En Joan pot Haver anat al cinema (57)
Since perfect aspectual markers preceding modals bar their epistemic
reading, the selectional restrictions of modals in such configurations must
be kept - Note that, as adjunct predicates, they do impose them. (33Ô)b.
and (341) below show that the selectional restrictions on the nominative NP
imposed by modal and infinitive must coincide:
(341) * Havia pogut ploure tot el dia (91)
Aspectual verbs have a parallel status as root modals although they
do not impose selectional restrictions on the nominative subject. They
nevertheless modify the infinitive they occur with; they have an adverbial
function. Note the ambiguity of (342)-(345): the aspectual verbs may either
have a motion verb interpretation - as main verbs -, or stand for a non-
predicative, adverb-like expression - future marker, iteration of action, etc.
In the latter sense they specify certain aspects of the predicate which
auxiliaries do not, and thus they are modifiers of the head.
(342) Anava a dir-te el que vapassar
(343) Be tomata felicitar-la
(344) Va venir a dir-me que no li interessava
(345) Ha arribat a molestar-la
The interaction of aspectual verbs - VP elements - and modals, like
aspect markers, indicates that root modals are VP elements. When
aspectual verbs precede a modal, the latter may only have a root reading
(346). An aspectual preceding douro - which may only have an epistemic
reading - is not possible (347). The reverse order is possible (346):
(346) Tornava a poder tocar oipiano (97)
(347) * Tornava a douro focar oi piano (99)
(34Ô) Bovia tornar a tocar oi piano ( 100)
Several other predictions follow from the analysis just sketched. One
of these is the fact that only the first modal in a sequence of modals may
have an epistemic reading. - Note that this is a difference between English
and Romance, as a sequence of modals is impossible in English-:
(349) La Núria doupodor tocar oi piano (101)
(350) En Jordi pot havor do soiucionar aquost assumpte moit aviat.
(109)
This is what is predicted if epistemic modals are analyzed as INFL
elements; there is only one slot in the structure for it. Note that (347) is
also explained by this: douro is an epistemic (only) verb, consequently
occupying the only INFL position in the structure; no other modal may
precede it because there simply is no position for it. If douro is followed by
a modal, the latter must be interpreted as a root modal. (351) is ruled out
because the selectional restrictions of podor are not met; in other words,
the selectional restrictions of the infinitive and the modal do not coincide.
(350 * Deu podar pJoiira
The fact that epistemic verbs in Catalan do not have infinitive or
gerund morphology, whereas root verbs do is another phenomenon
explained by this analysis, together with the asumption that the licensing
condition on epistemic modals is that they must be linked to morphological
tense. This is assumed to be true of all epistemic notions as possibility or
necessity, which must be relative to a given time. Licensing may take place
either under the relation predicate-argument ( "the modal constituting a
property of the time-frame expressed in the sentence"), or operator-
variable (" the value of the variable being determined by the features [ +/-
PAST ]").. The following illustrate the fact that if they have infinitival or
gerund morphology, modals may only have a root reading, in other words,
that epistemic modals must bear tense:
(352 ) Ignoràvem con? [PROpoder a&prassar-JJ aJ qua sant/am / ( 110)
(353) fPEO podant dadJcar-ta ajscJusivamant a also ¡no antaño
parqué no ño fas
(111)
This observation does not hold for English, a fact which may follow
from morphological constraints given the fact that English modals lack
infinitival and participial morphology.
Strozer (1961) *
Strozer proposes an alternative to the restructuring rule in Rizzi
(1902a) which implies a base solution. She argues that a base solution is
preferable and that, apart from theoretical problems that the restructuring
rule poses, there are also empirical inadequacies. The fact that there are
verbs that show a double subcategorization is taken as an argument in
favour of a possible VP-complementation; the assumption that, most
infinitives are clauses is not affected by the fact that a very limited set of
verbs take VP-complements. The following examples show that there is a
difference in meaning , and thus a plausible double subcategorization,
which in turn counts as evidence for a VP-subcategorization. The Italian
example does not display a difference in meaning and, thus, is assumed to
be a simple sentence:
(354) a. Ana volvió a f yp empezar a copiaria I
b. Ana volvió (allí) a is MP empezar a copiaria / (ô)(9)
(355)a. Siani eleve is presentarla a Francesco /
b. Gianni { y la óeve presentare fa Francesco
Strozer claims that there is no strong evidence to posit a bisentential
analysis for sentences like (355); that there are other reasons that may
account for the ungrammaticalities observed by Rizzi - cf. 2.3.2.1 -, which
he interprets as being evidence for constituency. An example:
(35b)a. çuesti argomentJ, a partarti del quaii verró ai più presto,
mi sem&rano moito interessant/'
b. *Çi?esti argomenti, apariare dei quale ti verro aipiüpresto....
The ungrammatically of this sentence could be due to the fact that
the trace of the clitic is not properly bound. (357) seems to be
ungrammatical due to restrictions on the breaking up of constituents by a
locative PP, a restriction which could also be at work in ungrammatical
sentences where Rizzi claims restructuring has applied (356):
(357) *Lo no a Firense messo ai carrante delia nostra dedsione (2 0)
(35Ô) a. Fra qualque giorno. verrò ad esport/Ja m/a ide a Firense
b. Fra qualque giorno, ti verrò ad esporre la mia idea a Firense
c. *Fra qualque giorno, tí verrò a Firenze ad esporre la mia idea
Thus, the tests used by Rizzi fail to provide evidence for
restructuring if other reasons are found that could explain the
ungrammatical examples.
What the VP-hypothesis shares with restructuring is that it also
applies to a small class of verbs , assumed not to be subcategorized for
clauses. The basic difference is that phrase-structure rules directly
generate the simple sentences in which restructuring complexes occur.
According to Strozer, the greatest shortcoming to restructuring is that the
cyclic transformation posited by Rizzi is too powerful in that it changes
structure throughout the derivation. The advantage of a base-generation
account is that no other extra device is needed.
220
2.4 An alternative theory: Guéron and Hoekstra (19ÔÔ )
Guéron and Hoekstra (19ÔÔ) (G&H) addresses the problem of verbal
complementation. They redefine the concepts of Jessica/ and auxiliary
verb and establish sets of verbal elements all characterized by their ability
or inability to mark their complements with a specific feature (a T-index ),
or assign them certain properties (a theta-role, or a T-role). Their
characterization of verbal elements makes crucial use of the notion of LF
and the fact that complements must be either, arguments or (parts of)
predicates. In their framework, complements are construed in LF as
nominal or verbal - i.e. as arguments or as (parts of) complex predicates. By
doing this, they formalize and extensively argue for the idea that auxiliary
verbs have verbal complements - VP - whereas lexical verbs have nominal
complements - NP, CP -.. The selection of a verbal complement is not
problematic; they draw upon the notion of sisterhood in Chomsky (19ô6b) -
cf section 1.2.2 - , where functional nodes do not count, and thus a V will
lexically select a VP eventhough functional nodes intervene. Proceeding in
this fashion, though, they come up with sets that differ substantially from
previous classifications - basically, they classify causative verbs as
auxiliaries, and modal verbs as having this option in certain languages -.
Their characterization is achieved by the introduction of formal
mechanisms ( T-marking, T-chain, T-role 58 )f which basically define the
circumstances under which a V may select a VP. The grouping of verbal
items as different types of verbs with very specific properties - some of
which may be subject to language variation - leads to a very constrained,
and compact theory on the complementation of verbal items.
G&H consider that the verbal or nominal nature of a projection is not
to be determined solely by the categorial nature of its head, but that
syntactic context plays a crucial role. An XP may be nominal or verbal
depending on the categories that dominate it .They propose the following
LF construal for complements:
(359) FimotionaJDetermination of Categories (FBC)
a. External
An XP is construed as a nominal projection iff it is casemarked
An XP is construed as a verbal projection iff it is T-marked
b. Internal
The subject of a nominal projection receives a Case which is
determined internal to XP; the subject of a verbal projection
receives Case ( if any ) determined by an external governor
(1)
Nominal projections are, thus, casemarked; their independence
follows from the fact that their subject is internally licensed - this includes
Genitive in NP and also Nominative in CP, as will be seen below - . Verbal
projections are not independent, they are part of complex predicates; they
are licensed by T-marking - Tense-marking - , which is a property of
auxiliary verbs only. This property is the formal instantiation of the fact
that auxiliary verbs, unlike lexical verbs, never denote an event, they
modify the event denoted by a VP. This event is situated in time by Tense,
and this is referred to by G&H as T-marking; the assignment of a T-index.
Lexical verbs may be T-marked in two ways: a) by Infi if it contains tense -
it is assumed that every CP introduces a new T-index ; and b) by an
auxiliary, if there is one. Auxiliaries may alternatively pass on the T-index
222
assigned by a higher auxiliary or by Infi. Lexical verbs absorb the Tense
value assigned to them and this becomes part oí the reference of the VP
which they head - and are the semantic head of -. Hence, the first basic
distinction between lexical and auxiliary verbs is that lexical verbs absorb
a T-index, and auxiliaries assjgn it. Note that it follows from the FDC above
that auxiliary verbs will always govern a verbal projection, and that lexical
verbs will not because they are not T-markers. Moreover, VP projections
that are not T-marked will not be licensed, and, therefore, will give rise to
an ungrammatical configuration. Although G&H do not explicitly make the
parallel, VPs are subject to a T-marking" Filter as NPs are subject to a Case
Filter. Only T-marking of a VP will identify it as a verbal projection at LF.59
Examples (36O) and (361) are not possibly identified as such because the
lexical verb croire cannot T-mark them. Examples (362) and (363) are
identified as verbal projections because an auxiliary verb governs them-
avolr in (362), laisser in (363):
(360) */e croyais Marie partir
(361 ) * Je croyais Marie aimée par Georges ( 17)f .,d.
(362) J'ai vu Pierre
(363) Eile laisse les enfants partir ( 19)
The following example instantiates how an XP may be interpreted as
nominal or verbal depending on the context:
(364) Eu lamento íjp os depiitedos terem trabalhado pouco]
(365) Eu lamento ÍXP1 teremj
ÎXP2 os deputados tj trabalhado pouco /
(7)a.,b.
The verb lamentar in Portuguese is assumed to optionally assign
case to its complement. From the FDC, if it assigns case then its complement
villi be interpreted as nominal; if it does not, as verbal. In (364), lamentar
assigns case to XP so it is nominal. As such, its subject must receive case
determined within the SP. This is the function of the inflected INFL in
Portuguese : it assigns Nominative case to the subject via subject-Infl
agreement. The structure of (365) is the result expected if we assume that
lamentar does not assign case and thus the XP must be interpreted as
verbal. This is assumed to follow from the fact that there is no Spec position
to the left of the head of the complement. If there is, the structure is as in
(364). As a verbal projection, the XP in (365) needs to be T-marked, but
since lamentar, a lexical verb cannot T-mark it, terem moves to Comp - or
alternatively adjoins to VP in LF - and identifies the complement as a
verbal projection. XP 1 and XP2 are analyzed as two segments of the same
projection because they share a head, terem - cf .Chomsky 1966b -. The LF
of (365) is :
(365*) £1? lamento Iypj terem 1ÍVP2 *>$ esputados ¿j/vpj
trabalnado peuco ¡jj (9)b.
where VP1= CP , VP2=IP and VP3=VP in the syntax; in LF they are
all interpreted as part of the same projection. The subject of the embedded
clause is assigned case under government by the raised auxiliary. The
identification of the subject is, thus, external - cf. (359), FDC - and not
internal by subject -Infi agreement as in (364). This is basically G&H's
account of Aux-to-Comp (Rizzi 1962c). The same factors apply to the
sequences that Rizzi considers in Italian:
(366) Suppongo [ ypj non esserj í yp2 Jui tj in grado di
affrontare Ja sititaztone / / / (14)
The lack of structures equivalent to (364) in Italian is assumed to be
the consequence of a difference in the infinitival morphology of the two
languages: Italian lacks a T-morpheme capable of assigning nominative
case. The lack of both of these structures in French follows if it is assumed
that French infinitival morphology also lacks this possibility, plus the fact
that the infinitival affix is too weak to assign nominative case to its own
subject by government, as is the case in (365") and (366). (367) contrasts
with (364), and (36Ô) with (365') and (366):
(367)* IJ regrette Jes députas avoir travaiJJepeu
(36o)* Je crois avoir Jes deputes travaiJJepeu ( 15)b.,c.
Several other basic assumptions are crucial in the framework of G&H.
Proper government is subsumed by antecedent government - proper
government is a function of chains (Chomsky 1966b) - and, thus, a link in
a chain will properly govern the link following it if there are no barriers
intervening. T-marking can void an XP of barrierhood (T-marking in G&H is
equivalent to L-marking in Chomsky (1906b)), so T-marking of a VP will
imply that the VP is not a barrier for an empty category in need of proper
government inside the VP - unless other factors hold, such as minimality,
as will be seen below . Note that certain consequences follow from this:
movement out of a VP is allowed - T-marking of VPs is a requirement -,
not out of a CP - a CP is never T-marked and case-marking is not identified
with L-marking. This is of crucial importance in their account of modal
structures and clitic climbing possibilities, as will be explained below.
The process of T-marking forms a T-chain : ( where k= the T-index
assigned by Tense )
(369) T* - [ aux* ] - [ vPk V* 1 (121)
The concept of T-chain provides a formalization of a specific
syntactic domain where the formation of a path provides a way for
feature percolation. Assuming that the external theta-role originates in the
V and is assigned to the external NP via INFL by subject-agr agreement,
the following example shows how theta-role percolates from the VP
through an auxiliary which is a link in the T-chain. This auxiliary is
assumed to have specific properties, as will be explained below. The
external theta-role reaches the NP-subject via I:
(3?0) John hasseen Mary
John {I fyp has Í yp laughed///
theta-1 (35)
There - structures and unaccusatives are instances of percolation of
Nominative case:
(371) There Js a problem
There fT-k fbe-k Í a problem ///
NOM (125)
Nominative case percolates through the T-chain and is assigned to
the postverbal NP ( it bears Nominative case because it determines the
form of the verbal inflection ).
(372) No sono arrivais molti
T&no i sono & / o j arrivaü'& ÍNPl n2oJtJ o/J
NOM (129)
As (372) shows, clitics climb within the T-chain, in the domain in
which features percolate. G&H show that clitic climbing and NP-raising are
possible within a T-chain because of the government and barrierhood
voiding properties of chains. A crucial fact about chains is that they may be
extended by Spec-Head agreement. G&H's account of passive relies on this:
(373) m I+tef ÍVPJ Vj ÍVP2 V2 i J//
John was killed
where i=j by SHAG (Specifier-Head Agreement) (74)
In such a structure, the trace of the raised NP is properly governed
by antecedent government. The verbal elements all share the same T-index
by virtue of the T-chain formed by T-marking. The chain is extended
because of SHAG which results in the proper government of the trace since
the other links in the chain share its same index^0 . Note that if V2 would
have another index by subject-head agreement within the VP, then
minimality would block the proper government of the trace - cf. (375) -.
The same process allows clitic climbing in structures where extended
chains are created:
(374) 11 lai est fidèle 11 j luij +1& [ yp-k est& ÍAP ti fidèle ej H (Ô3)a.
Note that the AP must be interpreted as a verbal projection - a
predicate -since it is T-marked by the verb., and as mentioned above, it
follows that movement out of verbal projections is allowed. The trace in
object position is antecedent governed by the adjective because it bears the
same index as its antecedent, and is, thus, a link in the same T-chain . SHAG
ensures the identity of k=i, and the adjective bears index i by subject-head
agreement internal to the AP. Note that the fact that the clitic attaches to I
ensures that the clitic and I share the same index - cf. also Chapter 3 and
Baker "s indexation convention on the X* dominating incorporated elements
-. As G&H point out, proper government of a clitic trace may be blocked by
minimality as in the following example:
(375) *U Jui/ croit [ Jean fidèle ej / (Ô3)c
Minimality blocks proper government of the clitic trace: the
adjective minimally governs it, but it is not coindexed with its antecedent
because it shares index with the subject of the AP.There is no raising of the
NP-subject of AP, so SHAG does not apply as a way to extend a chain,
contrary to what happens in (374) above. Moreover, the lack of T-marking
makes the complement an independent unit out of which extraction is not
possible - cf. the SSC -.
The fact that SCs may be IPs in the syntax and contain an INFL node
- i.e. the assumption that functional nodes are relevant in the syntax - ,
allows for the explanation of certain processes. As mentioned above, the
22Ô
external theta-role is assumed to be assigned by the VP via INFL. If INFL
absorbs the case assigned by the selecting lexical verb ,by visibility, it will
be allowed to "keep" the external theta-role, and to function as a
pronominal inflection. Causative structures in French exemplify this
phenomenon. G&H provide evidence to show that the complement of a
causative verb is , in fact, a VP, and thus the causative verb must be an
auxiliary ( cf. below):
(376)/<?/t?/a/ [jpl iyp lire ces livres // V-to-I
TH-1 TH-2
Je feral [ jp lirej [yp tj ces livres //
fifVPl ferai ÍVP2 llr#i ÍVPJ tl ces livres/ff IS
TH-1 TH-2 (17Ô)-(179)
Linked with the fact that lexical verbs are the only verbal elements
that may have nominal complements is the fact that they are the only ones
that may assign theta-roles. An XP is nominal only if it is assigned a theta-
role. Auxiliaries do not assign theta-roles, but they may assign Tense-roles
(T-roles)D 1 to their complements. Auxiliaries which assign T-roles select a
VP with an independent tense. This creates a bifurcation of auxiliaries into
two different types: T-auxiliaries ( those that assign T-roles) and neutral
auxiliaries ( those that do not assign T-roles). T- auxiliaries govern a VP
with an independent Tense-morpheme. Neutral auxiliaries combine with
the Tense-morpheme of their complement to form a complex tense and
define the tense of S. BE is assumed to be a T-auxiliary; HAVE a neutral
auxiliary. T- auxiliaries and Neutral auxiliaries show systematic properties
that justify their classification into two distinct classes. One of such
properties is that only neutral auxiliaries allow the percolation of the
229
external theta-role assigned by the VP to the subject. T-auxiliaries do not.
T-auxiliaries, thus, involve Raising or Control whereas neutral auxiliaries
involve VP complemements with no subject position and direct percolation
of theta-role to matrix subject position. (12) above illustrates percolation
over a neutral auxiliary, Mve. A passive structure like (15) illustrates no
external theta-role percolation over the auxiliary: the external theta-role is
assumed to be assigned to the verbal affix of the participle. The causative
structure (18) also illustrates a T-auxiliary structure where all theta-roles
are assigned within the complement of the auxiliary - via the pronominal
Infi.
The observed correlations between T- role assigning properties of
auxiliaries and thematic relationships, make G&H unify them under a
specific notion, Complete Thematic Constituent:
(377) An XP which is assigned a T-role is a Complete Thematic
Constituent (CTC): all theta-roles associated with X, the head of XP,
are assigned internal to XP (34)
This also captures the idea that tense is an operator on a theta-
domain. The identity of both domains is illustrated by a number of
syntactic processes. Note that precisely the examples mentioned to
illustrate theta-role percolation/ non-percolation illustrate the concept of
CTC: the complement of a passive (373) is a CTC: all its theta-roles are
assigned internally- A? is a T-auxiliary and assigns a T-role - ; the
complement of a causative (376) is a CTC: all its theta-roles are assigned
internally- faire is a T-auxiliary and assigns a T-role-. The complement of
the auxiliary A?r<? (370) is not a CTC: it is a neutral auxiliary ; the external
theta-role is percolated onto the matrix subject position, and the resulting
tense is complex - a complex past- as a result of the combination of both
Tense-morphemes.
The grouping of auxiliaries into two different classes and the
postulation of constraints based on their properties, accounts for a problem
essentially left unexplained in previous analyses: the fixed order of
auxiliaries as well as their distribution - cf. section 2.1, and 2.2.2. G&H
postulate the following two constraints:
(37Ô)a. A T- chain may not contain two auxiliaries of the same class
b. Neutral auxiliaries precede T-auxiliaries (40)
If auxiliaries are assumed to have a function in terms of assigning a T-
role - only T-auxiliaries -, then a. follows from the assumption that there
should only be one auxiliary with a specific function in a T-chain. - in other
words, that auxiliaries with the same syntactic function are in
complementary distribution -. If theta-role percolation is only allowed by
neutral auxiliaries then b. follows because the opposite order is impossible:
a T-auxiliary would block the external theta-role percolated through a
neutral auxiliary on its way to the matrix subject position. The following
examples illustrate violations of (37Ô)a. and (37ô)b. :
(3?9) * ft fais avoir travaillé Marie b.
T-aux/neutral (25)b.
(36" 0) * John wast nave seen b.
T-aux/neutral (41 )a.
(3Ô1) * Mary made John have played with the baby b.
T-aux/neutral (41)b.
(3ô2) * MarJoJoavrebbe doruto aver finite a.
netural/ neutral (2l6)b.
(3Ô3) * M est été vu a.
T-aux/T-aux (27)a.
(3Ô4) * fíe is been seen a.
T-aux/T-aux (27)b.
The characterisation of auxiliaries is subject to language variation, as
will become clear in the summary of G&H' s account of modal complement
structures. This characterisation, though, is also subject to variation within
the same language; the same lexical item may be a neutral or a T-auxiliary
depending on the context, as is the case with have in English:
(3Ô5) John has burned his house Neutral aux
(366) fthn had his house burned T-aux
Causative and modal verbs:
G&H classify Italian, French and English causative verbs as auxiliaries on
the basis of three basic facts: they can intervene between a raised object
and an extraction site (367) and (3ÔÔ) - only auxiliaries "continue" a T-
chain, and, thus make the proper government of the trace possible - ; they
can bear the value of tense required for VP anaphora (3Ô9), which is
assumed to be a property of auxiliaries ^2 - cf. section 2.1, no. 6 -; and they
take bare infinitives (32), which are assumed to be VPs as opposed to "to-
infinitives" -:
(3Ô7) ft l'ai fait voir
(3ôÔ) llihrifurono fattileggere ( 173)a.,b.
(3Ô9) Jean a acheté du pain ce matin et Pierre le fera ce soir ( 174)b.
(390) They let John leave (175)a.
English make /let, and French faire are structural case assigners so
they may case-mark the subject of the VP complement. Note that by being
T-auxiliaries, they assign a T-role to their complement and give it the
status of a CTC, whose external theta-role is assigned within its VP-
projection. :
(391) We made/let ¡the warden hang the prisoners /
(392 ) Nous avons laisse {le gardien prendre les prisoniers I
(176)b.,c.
For French faire, as explained for (376), case is absorbed by INFL
and the external theta-role is assigned to it. As (393) shows, clitic climbing
is possible in causatives in French, as follows from the assumption that they
are auxiliaries, they take a VP, and that extraction is only possible out of a
(T-marked) VP:
(393) ft T& les/ ferai* [lire-*' ej/ (163)
Modal verbs are assumed to be always auxiliaries only in English -
not in French, nor Italian -. Root modals can function as both, auxiliaries
and lexical verbs in Italian and French. In French they raise to INFL in
infinitival structures like auxiliaries and unlike lexical verbs - cf. Pollock
(19Ô7), and Chapter 4, sections 4.2,and 4.3 -. The following contrast seems
to imply that they are T-markers:
(394) a. ? Il pensait ne pouvoir] f yp pas tj dormiria'/
b* 11 pensait ne croirej [yp pas tj auxnistoiresdefantomesj
(191)d.,f.
Their ability to allow VP anaphora is also evidence for their auxiliary
status:
(395) a. Pierre voudrait accorderJe piano maisilne peutsYeut/doit
pas
b* Pierre voudrait accorderJe piano maisilne décide/croit pas
(192)
In Italian, there are also two possibilities. Nevertheless, as Italian
allows clitic climbing in modal complement structures, the different
function of the modal verbal element may be contrasted with respect to
this structure, which is not possible in French. When modal verbs are
lexical, they assign case to their complement identified as a nominal
projection. The subject position is, thus, not case assigned from the outside,
and the external theta-role of the embedded verb is taken up by PRO in
Spec-IP position. It is a Control structure:
(396) Mario vuolel' ¡p PRO / / • [ypleggere il libro III
TH-1 TH-1 (193)t>.
If the modal functions as an auxiliary, the case that it assigns is
absorbed by AGR in INFL, which then may bear the external theta-role of
the embedded verb. The pronominal argument inflection is subject to
Control from the matrix subject. The auxiliary assigns a T-role to its
complement and the CTC requirement is satisfied:
(397) Mario vuote [ yp logg&rej iyp tj iiiibroi]
TH-1 TH-1 (194)b.
Clitic climbing is only allowed in a structure like (397) because the
trace of the clitic is a link, in the unique T-chain which includes all the
verbal elements in the sequence. The embedded verb properly antecedent
governs the trace, as was exemplified above in the French causative
structure (376):
C39Ô) Mario Joj + vuoio & {leggerek oj] (196)b.
Clitic climbing is not possible in a structure like (396) because there
are two T-chains in the structure; the lexical modal verb does not T-mark
its complement, so each of the subjects is the subject of a distinct T-chain.
This invalidates a structure like the following:^
(399) Mario 1* Jo/ + vuoJo ¿' [PRO //•* [Joggoro ¿ o/ J/J ( 195)b.
G&H explain the fact that French modal complement structures
equivalent to the Italian structures do not allow clitic climbing by making
reference - following Kayne (1967) - to the properties of AGR in INFL: only
the AGR of pro-drop languages may function as a syntactically active
pronominal argument. Note that the fact that it functions as a pronominal
inflection is not only a characteristic of pro-drop languages - cf. French
causative structures (376) -, what is is the fact that it is a pronominal
argument subject to syntactic processes such as Control, which is assumed
to be the case in (397), as implied by the fact that the infinitival and the
matrix subject have the same theta-role. This process is assumed to be
what defines INFL as an A-position; French needs the Spec-IP position. The
lack of such a possibility explains the lack of structures like (397) and
(39Ô) in French.
This account is problematic for English , which is not a pro-drop
language and its modals take a VP complement. G&H attribute this to the
fact that there is a strong verbal affix in modals, which may function as the
external argument of the VP it governs. In (400), the modal assigns its
external theta-role to the matrix subject, while it takes the external theta -
role of the embedded VP - its verbal affix having inherent case -:
(400) Johßj {must iyp examine hispatients/
TH-1 TH-li TH-2 (203)
Epistemic modals in English are also auxiliaries , in French they are
lexical verbs, and in Italian they may be both. The contrast between (401)
and (402) shows that they are different in French and Italian:
(401) * Jeaniliaj devrait*[ P&Oj JP voir? tj (206)b.
(402 ) Giannij iioj deve J [ agr vedere J' ej/ (2 05)b.
The French modal assigns case to its complement; it is interpreted as
a nominal projection with an internal subject. Raising would be blocked by
minimality, as in the Italian structure (399). In the Italian example, (402),
AGR in INFL absorbs the case assigned by the modal and functions as an
expletive pronoun. Proper government of the trace is not blocked because
vedere, the nearest link in the chain shares indices with the antecedent of
the trace by virtue of the fact that the matrix auxiliary, which bears the
clitic, T-marksit.
The avere -» essere rule
The non-rigidity oí the properties oí verbal items - i.e. their ability to
have them or not - is basic in G&H' s account oí the avere— essere rule
which, as has been shown in section 2.3.2.1 , applies in clitic climbing
structures after restructuring:
(403) a. Maria Jfa volute venird
b. Maria d è yoJuta venire ( 169)
It is assumed that Italian essere as opposed to French être has the
possibility of refraining from assigning a T-role , but still preventing theta-
role percolation. There is, thus, a neutral essere in Italian. This property of
essere is illustrated in several processes some of which draw upon the
theory internal restrictions on the order of auxiliaries. Neutral essere may
precede a T-auxiliary, as opposed to French être:
(404) a. Mar/a éstata invitata
b. Marie a été invitée * (I65)a.
(405) a. Jlibrifuroso fattiieggere
b. * Leslivresfurent / ont été fait, lire ( 165)c.
Italian stare is a T-auxiliary which may be preceded, not followed,
by essere.
(406) Fssendo Mario state tortiirato .... ( 166)b. (407) * Mario stava essendo bastonato ( 167)
The account of the rule of avore — essore draws upon the
assumption that modals are T-auxiliaries in clitic climbing structures, and
the fact that essere may be a neutral auxiliary- and thus precede a T-
auxiliary - while not allowing the percolation of the external theta-role
assigned by the VP. The complement of the modal is a CTC, and all theta-
roles are assigned internally. It is a raising structure:
(406) Mariai fvpj ci/ è[yp2 e/ voiuta / ypj venire ej/fe//
(170)b.
In non-clitic climbing structures, the choice is avere, which allows
the percolation of the external theta-role when the modal is a lexical verb.
The complement of the modal is a nominal projection. It is a Control
structure:
(409)Mariay/vp/ A? /VP2 volute lo> PSOj venird e/ ///(170)a.
In French, there is no auxiliary change although there is a marginal
structure where clitic climbing is permitted ( Kayne 1967):
(410) ? Jean y a voulu aller
(411) *Jean y est voulu aJJer (172)
This follows from the difference between £¿rt? and AÏS*?/*? ; the
French auxiliary is always a T-auxiliary, thus, the only auxiliary which may
precede other auxiliaries in French is avcdr.
2.5 A few remarks
The inevitable changes in the model have led to the invalidation of
some of the mechanisms used in the sketched proposals. In other words,
pre-GB proposals are often invalidated because they imply a violation of GB
principles. This is the case, as already observed, with the first attempt to
explain the peculiar nature of "half-blooded" verbs by restructuring , which
was deemed no longer tenable as it led to a violation of the Projection
Principle, modifying the lexical properties of the verbs throughout the
derivation. The same can be observed for the first attempt to explain the
behaviour of causative predicates; i.e. Kayne (1975) - as noted in section
2.3.I.O -. Burzio (1966) discusses the proposal - cf. 2.3.1.2 - and rephrases
aspects of it. Quoting Burzio: " This is an obvious weakness, given Kayne's
extensive and convincing discussion of the similarities between passives
and FP. This weakness arises from the inability of the ST framework to
separate the three properties of passives, as is rather clear from Kayne "s
own discussion " (p. 256) (italics mine).
Nevertheless, Burzio (19ô6)'s VP movement analyses are also
dubious. Although they are independently motivated for certain
constructions - cf. 2.1. and preposing structures -, there seems to be no
direct motivation to posit such a mechanism in either C- or M-A sequences.
The basic proposal in this work necessitates several ad hoc mechanisms
such as " Subject Substitution " - cf. section 2.3.2.2 -; does not provide a
trigger for VP movement; nor does it establish and license the exact landing
site for the moved VP. See Chapter 3 - section 3.3.2 - for another VP
movement analysis - at LF - by Baker (190*0) and - section 3.4 - for some
criticisms of this proposal.
240
VP complementation íor complex predicates must confront strong
arguments in favour of a clausal status of tue node dominating the verbal
complement - cf. especially section 2.2.1 -. It will be explained in section
3.4 - see also 2.4 - how the alternative theory proposed by Guéron and
Hoekstra (19ÔÔ) allows for a preservation of clausal complementation of
most instances of Vis in complex predicate structures in the syntax,
while it construes them as VPs in LF. In this fashion, it reconciles
VP complementation proposals and the strong syntactic clausal
complementation arguments.
The simultaneity of two parallel structures must face the assumption
that principles hold at different levels but both structures are present at all
levels. Principles requiring a monosentential structure - such as Binding -
will obviously be violated by the other, bisentential, yet simultaneous
structure. If mechanisms are found that avoid this - inevitably undesirable
- consequence, I believe they should be assumed.
Adjunctions to V proposed for adjunct predicates- as in Picallo
(1985) - are not independently motivated, and shown inviable in Chomsky
(1966b) - cf. i.2.2 -: adjunctions are only allowed to Xo or XP, not to X'.
Although, as noted in section 2.3.2.3, Picallo (1990) adjoins root modals to
VP, and not to V. The proposal for root modals differs from the hypothesis
for epistemic modals, for which a VP complementation seems to hold, and
for which I will tentatively argue for in Chapter 3.
Other mechanisms such as thematic rewriting rules and reanalysis
may be accounted for in the light of the present theory in ways not
exclusively needed for these processes. Such ways include chain formation,
- which may apply to verbs in a sequence as shown by G&H -, and head
movement. Together with these, one might explore the importance of LF as
a way to explain the peculiar behaviour of some verbs in a sequence - in
241
line with G&H, as mentioned -. The possibility of consecutive VPs - cf.
section 3.3.I.2- is, at present, not necessarily problematic to the principles
of grammar.
Explanatory adequacy - and, thus, independent motivation for the
chosen mechanism - will be the guiding lines to the approach adopted in
Chapter 3.
Before bringing in the main hypothesis in the thesis, though, I would
like to point out some other ( apparently ) problematic facts which arise in
view of the proposals sketched for complex predicates. These are not
theoretical but empirical facts from Catalan and Spanish. There are verbs
which are not traditionally classified under M-A nor C-predicates and
which, nevertheless, allow clitic climbing. The folllowing illustrate this
phenomenon:
(412 )a. Ho semblava entendre tot
b. L "na aconseguit seduir
c. L "ña decidit trucar
(413)a. La intento convencer
b. Lo deseaba ver
c. Lo pretendió ridiculizar
It must be noted that not all speakers find these fully grammatical,
but the fact that most speakers do, indicates that whichever mechanism is
found valid for the formation of complex predicates should apply to these
verbs as well. 6 4
Notes to Chapter 2
242
(1) I am deliberately leaving aside important studies on the matter which
do not follow a generativist view. This has been done in order to keep a
coherent presentation of the issue. Moreover, the aim of this thesis is not to
compare or contrast frameworks but to put forward an attempt of
explanation within one specific framework, as the title implies.
Observations made by non-generativist authors such as Palmer ,
Huddleston, Quirk, etc. have, at certain points, been included.
(2) Môdàîb in English M ve been traditionally assumed to be a distinct
category M and, therefore, generated under a different node Aux or INFL.
This issue will be touched upon in several sections in what follows .
(3) In the sections where a specific proposal is summarized, the number
next to the example indicates the number in the source.
(4) There are other instances in the language of this same phenomenon
with other verbs: soy, estoy.
(5) Although Zagona (19ÔÔ) proposes a framework where auxiliaries
assign specific theta-roles, temporal roles - cf. section 2.2.2 -.
(6) See Chapter 4, section 4.1, Abney (1966), Fukui & Speas (19ÔÔ) for the
use of this argument to grant elements a functional status. A question which
arises within the present framework" and which is briefly touched upon in
Chapter 4 is whether auxiliaries may be generated in funtional nodes.
(7) In languages with other choices for the head parameter, this order may
be reversed; e.g. Basque: Maiteak Jon ikusd du ; where ikusi is a non-
finite form of the main verb (to see) and du is the auxiliary ( 3sgS - 3sg0).
(Ô) See sections 2.4 and IA for an attempt to explain the ordering and the
distribution restrictions of auxiliaries within the present framework..
(9) The Affix Hopping Rule is valid for English sequences of auxiliaries.
Modals in Romance languages differ from modals in English and, as will
become clear in later sections, they are not traditionally considered
auxiliaries. Even in recent proposals which do consider them auxiliaries - cf.
2.4 -, they are only so considered in certain configurations. See section 2.3.2
for the debate on modals in Romance. In 2.1.2, modals and aspectuals are
regarded as having a double nature, the fact which all sections summarized
which deal with M-A sequences try to account for.
(10) Note, though, that British English savo allows for constructions of the
type: (i) savo you asy eslieron?. Here, HAVE cannot be considered an
auxiliary. See Pollock (19Ô7) for an explanation in the present framework for
this special behaviour of British English HAVE.
( 11) It must be noted that in the present framework, the concept of adjusct-
prodlcate has been posited - cf. Zubizarreta (19Ô2), Picallo (1985X0 990)
among others -, which would allow a verb imposing a specific type of
selectional restriction not to be a main verb - cf. section 2.3-2.3 -.
(12) I will use the term clitic climbing as a general term , bearing in mind
that the phenomenon is far from being established. The original work
referred to in this section relied on a movement account of clitic placement
because me model tnen used couia account ror its distribution on me basis
of the Specified Subject Condition (SSC), a condition on transformations
which disallowed movement if there was a subject intervening between the
extracting site and the landing site - cf. also section 2.3.2 -. An alternative
account is to consider clitics as base-generated on me verb (or INFL) and
licensing an empty category in their thematic position. - cf. Borer (1964),
Burzio C1956), Rosselló ( 1966), among others.
(13) There are important restrictions with respect to different types of
adverbs -cf. especially jackendoff (1972) -, a type of evidence which I have
not considered in this thesis as adverb typology is not, to my mind, directly
related to its main proposal.
(14) See Hernanz&Rigau (19Ô4), Belletti(19ô2) among others, for discussion
on this issue.
(15) Examples like (i) and (ii) show mat not all main verbs disallow clitic
climbing in either Spanish and Catalan. See section 2.5 for some
observations on mis fact.
(i) Ho va iieJjsar veure ajs seus antics
(ii) la Mentó ver
See also note (64)
(16) Note that the examples in (g) correspond to an aspectual verb and a
modal verb. H&R obviously refer to me contrast shown by auxiliaries as in
test 6, which disallow anaphora - cf. Chapter 3 for an explanation of mis
constrast and main proposal of mis thesis.
(17) Note also that most main verbs do not allow Null Complement
Anaphora, as pointed out by ]M Brucart -p.c.-: (i) *Lius entró finalmente a
Ja fiesta, pero ne deseaba. The fact that the Spanish and Catalan examples
in test 6 which are used to illustrate that main verbs do allow VP anaphora,
include a clitic; a fact which is obviously not to be disregarded in any
explanation oí VP anaphora - which is beyond the aim of this study.
(10) Note that there are other structures where two verbs are also
consecutive : En Eamon es dutjsava cantant / Corría cantando / He left
laughing at us .These are accounted for in the literature by positing either
a secondary predication or a small clause (cf. Chomsky (19Ô1), Stowell
(19Ô3), Williams (19Ô0) among others). Note that these structures often do
not imply a sequence of two consecutive verbs: fíe left the room laughing
at us/ En Eamon esdutsava cada dia cantant /Pepe corria los maratones
cantando. Thus, I will not take these sequences into consideration.
(19) These are not the only possible control structures. I will not consider
those which do not involve a sequence of verbs, for instance, interrogative
infinitives as: / wonder who PEO to ask to the party or cases of non-
obligatory Control: PEO to leave the city during the Olympics would be a
verygoodldea.
(20) In using the distinction suntactle / lexical passive, I follow Wasow
(1977). By syntactic passive I refer to the fact that the NP which has been
raised to subject position is not related to the main verb in the lexicon; i.e.
it is not one of its arguments. In a lexical passive construction, the NP
which has been raised is the argument of the matrix verb, as in : She lost
the key / The £ey was lost. In the Catalan and Spanish examples, the
impossibility of syntactic passive implies the raising of a non-argument of
tne matrix vert) - creure /creer-; i.e it is tne tue external argument or the
embedded verb which is raised.
(21) Nevertheless, there are alternative approaches within the present
model ; approaches which postulate a base-generated VP node for some
infinitival structures. Some of these will be discussed below - section 2.3.-.
There is an important alternative within the framework: Guéron and
Hoekstra (19ÔÔ), who grant some infinitival structures a VP status at LF -
cf. section 2.3.3.
(22) As already noted in Chapter 1, in the summary of Control Theory, the
status of PRO has been the issue of much debate. The so-called PRO
Theorem has been argued to follow from other subtheories, thus, the effect
of Control to be reducible to other modules. See Manzini (1963a), Borer
(1969X Bouchard (19Ô4) among others.
(23) Note that for- deletion is not allowed after N or A, only after V: * It is
silly you not to call Mm / * Tne desire you to finish your thesis.
(24) Nevertheless, in the present framework, INFL may be granted certain
properties which allow it to function as a pronominal - cf. Guéron &
Hoekstra (19ÔÔ) -, or verbs may be linked in other ways - cf. Picallo (19Ô5)
, Zubizarreta (1965)- The arguments alluded to in this section, thus, lose
strength.
(25) But see section 2.3.2.3 for some'monosentential analyses of complex
predicate structures within the present framework.
247
(26) Zubizarreta (19Ô5) uses the term complex verb for a sequence of
causative plus main verb; a sequence for which I use the label complex
predicate.
(27) I include modals in the examples for completeness, but as has been
observed - cf. Chapter 1 section 1.1 -, they are assumed to be generated in
a different node from other auxiliaries- cf. also Chapters 3 and 4 -.
(20) The reason why a full verbal sequence ( 5 verbal elements ) is not
grammatical in Catalan nor in Spanish as in (i) and (ii) respectively, calls for
an explanation:
(i) *? Els caramels que no trobem poden haver estat essent menjats per les
criatures
(ii)*? Los caramelos que no encontramos pueden haber estado siendo
comidos por los niños
This thesis deals with 2 verb sequences, so the account of this restriction is
beyond its scope. Note that the English examples are not fully acceptable
either.
The fact that the modals in the sequences in the examples in the text are
interpreted as epistemic will gain importance in Chapter 3 - cf. also 2.3.2.3-
(29) As noted in the introduction, in this thesis Í will concentrate in the a.
sequences - leaving progressive and passive for further research.
(30) Functional nodes have been briefly introduced in Chapter 1, and will
not be further taken into account until Chapter 4. In this section the issue is
to consider the structure of complex verbs bearing in mind that all verbal
elements are members of the V category, but disregarding - for the sake of
simplicity - tüe important matter oí whether they should be granted a
distinct, functional, node - an issue to be touched upon in section 4.3 .
(31) I am leaving aside comment on other studies referring to ASW's
proposal which are not within the present model. Note the article by
Gazdar, Pullum, and Sag (19Ô2).
(32) I will not make crucial use of Zagona*s framework in the rest of the
thesis. More precisely, I not consider part of her framework, although I will
refer to it in several occasions ; i.e. the licensing of VPs . I will assume her
structural conclusions. The licensing of empty and full VPs is, obviously, a
very important matter, but beyond the scope of this thesis. It must be
noted that - to my knowledge - there are two important proposals on this
subject in the field: one is Zagona (198Ô), the other is Guéron&Hoekstra
(i960). The latter will be considered in 2.4 and in Chapter 3 as it touches
upon not only auxiliary constructions, but also complex predicates. Zagona
(1956), nevertheless, provides the structure necessary to postulate V-0
movement in Chapter 3-
(33) I will not consider this important matter, again because it lies outside
the scope of this thesis. Nevertheless, section 2.4 refers to Guéron &
Hoekstra (I960)' s framework who, using Tense as an unanalysed primitive
provide a basis for an account of the distribution of auxiliaries which seems
to me not at all implausible. G&H also directly avoid some of the theoretical
problems observed below as a consequence of the subcategorization
licensing condition on VPs.
(345 An important explanation of this fact is provided by Pollock (19Ô7)
relying on the parameterization of inflectional nodes - cf. Chapter 4, section
4.2-.
(35) Zagona does not refer to Baker's framework of incorporation , although
V-to-V is also seen as an instance of head-movement as in Barriers.
Zagona's claim of participle incorporation is related to the main hypothesis
in this thesis, as will be posited in Chapter 3, although it crucially differs
from it because I do not consider temporal-role marking.
(36) The status of X is left undetermined here, as this is the matter of
debate among the scholars in the literature sketched in the following
sections . Nevertheless, as has been made clear in section 2.2.1, there are
many reasons to consider X a clausal node.
(37) The capitalized English verbs are used to refer to causatives in general.
When a distinction is needed between English and a Romance language, it is
made by using the corresponding lexical items. This also appiles to other
sections.
(30) The term reanaJysis is defined differently in Manzini (1903t»), section
2.3.1.4.
(39) Brucart (1984) assumes a process of restructuring proposed by
Bordelois (1974), which I do not consider in the review of proposals.
(40) See Chapter 4, section 4.3 for a different account of Past participle
agreement in the present framework; cf. Kayne ( 19Ô7)
(41) Note that structure such as (215) in the text implies allowing for a VP-
internal subject. Nevertheless, Manzinii 19o3*>) does not refer to the present
debate on an original internal position for the subject - and subsequent
movement to the specifier of a functional projection - because it was
posited in later works - cf. Chapter 1 ( p. 11) for references.
(42) The fact that structures like J'allait une table exist, imply that faire
may also assign Case. This according to Manzini must, be captured in the
lexical entry by an implication: if the verb assigns a theta-role to a small
clause, then it must be a reanalyser.
(43) It must be noted that clitic climbing is allowed in some control
structures, verbal sequences composed of a main verb not characterized as
a modal plus an infitive, both in Catalan and in Spanish. This is briefly
touched upon in section 2.5, where some examples are given. See also note
( 15) and last note to this Chapter.
(44) In Catalan there is always agreement with the modal verb, regardless
of the position of the object:
(i) Està segur que espoden aturar fes situations injustes
(ii) Fstà segur que ies situacions injustes es poden aturar
(iii) *£sta' segur que espot aturaries situacions injustes
(45) Note that not all the verbal elements in (241) conform to all of these
properties: dare/need are also non-modals, is to has all the finite forms.
The characteristics of modals in English are not considered in depth as they
do not bear directly upon the subject of this thesis.
(46) It must be noted that this distinction - due to Hof mann (197o) -
disregards other classifications of modality into more different types as in
Palmer (1979). Palmer makes a difference between epistemic modality -
" making a judgement about the truth of a proposition" ; deontic modality -
influencing the action as by giving permission, or imposing an obligation,
and dynamic modality - which refers to ability or volition of the subject. I
will not consider this classification - nor others - but follow the general use
in generative studies which unify deontic and dynamic into "root" . The
subject of modality types is beyond the scope of this thesis.
(47) It must be noted that the second examples allow an epistemic
interpretation: (i) Us governs poden posar fi a la injustícia, però no volen
fer-ho. In the examples I intend a root interpretation; i.e. (ii)The
governments have sufficient power to put an end to injustice".
(4ô) In Spanish the distinction is made by including the preposition de
after deber when intending an epistemic interpretation, as the parentheses
show. I have included the parentheses because this is not followed by all
speakers.
(49) The syntax of aspectual verbs is more complicated than that of other
"restructuring" verbs as they have several other options: they may be
followed by other non-finite forms, and, indeed, there are some aspectual
verbs that must be followed by an -ing form, while others allow both:
(i)a. Continua *fer/fent les mateixes classes
b. la policia comença a desea/regar / ¿descarregant quan rep
ordres
(ii)a. Will Israel ever stop *tc kill /killing Palestinians
b. Will Israel ever cease to kill /killing Palestinians
Note that the relevant interpretation of (i)b is what makes the second
option ungrammatical, the verb comencarm&y be followed by an -ing form
if interpreted as in (iii):
(iii) Els actors comencen J "obra Interrogant eJpuMc
(50) Catalan and Spanish have ways of disambiguating : the use of a
different preposition or a periphrastic construction :
(i) Ha tornat per CtaJde JdJr-JIque....
iii) Ha vlngutper {taJde)dirque...
(iii) Ha venido para decirme que...
(iv) Ha vuelto para dearie que... )
(51) As pointed out to me by ]M. Brucart, this test is also applicable to
some modal verbs, as in (i) for Spanish and (ii) for Catalan:
(i)a. Luis debe de estar enfermo
b. Luis está probablemente enfermo
(ii) a. En Lluís deu estar malalt
b. En Lluís està probablement malalt
These modals are interpreted as epistemic - as already noted deiwe
may only have an epistemic reading in Catalan, cf. also note (46*).
(52) It must be mentioned that several important analyses have been left
out in this section. I am basically referring to the parallel structure analysis
in Zubizarreta (1962) and the "parallel phrase-marker" analysis in
Manzinii 19ô3b). The former follows the same proposal explained in section
2.3.I.3. Manzini incorporates restructuring into a theory of phrase-markers:
she allows a double phrase-marker for M-A sequences one of which may
be derived from the other.
(53) The NP subject of the complement clause is PRO, but the fact that there
is a subject position does not, in Rizzi's framework, prevent the rule from
applying . The present framework disallows such a "destruction" of
structure. Linguists who have posited a non-sentential status for the
embedded clause have had to somehow rearrange the dischargement of the
subject role - cf. especially Guéron and Hoekstra (I960) -.
(54) Note that it also applies vacuously if the clitic takes a "long step". The
argument is theory internal; it relies on the assumption that there Is- a
clausal node.
(55) Note that agreement with the verb and postverbal occurrence of
subject ( preposed object ) are possible. This is regarded as a consequence
of the available post-verbal position for subjects in pro-drop languages.
(56) I have only included examples illustrating clitic climbing because
object preposing and auxiliary choice - cf. also 2.1 - are not relevant tests in
Catalan nor Spanish. See Rizzi (1902a) for illustrations on these
constructions for Italian.
(57) It must be noted that in Picallo (1990) the same proposal is made, but
root modal verbs are generated as VP-adjuncts - (9), p. 259 -.
(50) It must be noted, as will become clear in what follows, that G&H*s
concept of T-marking, and of Tense-role, is not equivalent to Zagona's - cf.
section 2.2.2 - notion of temporal-marking. She makes use of distinct
temporal roles - E,R,S -, which are not considered by G&H. Their notion of
T-chain is also distinct from Picallo(19Ô5)'s use of T-chain as a redefinition
of a governing category of an element.
(59) G&H do not make the parallel of T-marking and theta-marking in the
same way that Zagona(19ÔÔ) does: the assignment of Tense-roles by
auxiliaries - see below - does not lead them to a theory of "argumenthood"
for VP projections, as is clearly expressed by the FDC ,(359) in text. Their
theory avoids interaction problems with NP/CP argumenthood - basically
VP-adjunction -, some of which were noted in sectioon 2.2.2 -. See also note
(50).
(60) The mechanism of chain extension by Spec-head agreement is also
crucially used in Barriers to account for passive in English, and also in
Zagona (19ÔÔ), following the barri&rs explanation.
(61) See note (59).
(62) Note that there is a reversal in the use of this test as compared to its
use in 2.1: VP anaphora was assumed to be a characteristic that only main
verbs could have in Romance languages. By considering causatives as
auxiliaries, G&H generalize the English test to Romance. Obviously, though,
the presence of the clitic cannot be disregarded.
(63) The fact that there exists the possibility of clitic climbing with other
types of verbal sequences has already been pointed out in note (15), and
will be briefly touched upon in section 2.5. See note (64).
(64) I must note, though, that the main proposal in this thesis refers to
complex verbs . Nevertheless, and because of the fact that they are
verbal sequences, a hypothesis will also be posited for complex
predicates in Catalan and Spanish; namely Guéron and Hoekstra (19Ôô)"s
T-tlieory proposal. Hence, the feature that the verbs in the examples would
acquire would be that of optionally being T-markers - cf. sections 2.4, 3.4,
4.3 - . Further research is obviously required; i.e. the specific restrictions on
which type of verbs allow this optionality. The fact is that not all main
verbs do, as there are clear ungrammatical examples of clitic climbing with
two main verbs in both Catalan and Spanish as in (i) and (ii) ( Í10) in H&R
1954):
(i) * Hi Jamento pensar
(ii) *la Jámente pegar
CHAPTER THREE: Incorporation
3.1 An outline of Morphology Theory
3.1.1 A note on non-evident single constituents
The observations in this section are intended to show basically that it
may be the case that not all morphologically separate units are also
syntactically separate. This approach is , to my mind., legitimated by the
kind of theory of morphology that will be presented in the next section. 1
Certain verbal sequences of two verbs are not evident single
separate units because they function as one lexical unit - they cannot be
divided under any circumstances, and it is often the case - in certain forms
of the tenses - that the first part of the sequence never bears stress -. I
believe it is worthwhile pursuing the idea that these do constitute a unit,
even though they do not surface as one morphologically complex form. In
other words, we may try to characterize their behaviour by assuming that
they become an Xo by the application of a syntactic process; namely, head-
movement, which allows for further cliticisation of the auxiliary verb onto
its host. In section 3-3-1 I will present the syntactic analysis and in the
next section I will review the theory of morphology assumed in Baker
( 19ÔÔ) - and Ouhalla (I960) among others - which seems to be relevant for
the explanation of the behaviour of such sequences.
Before presenting the theory of morphology which may lead to an
explanation of such a phenomenon, it may be wise to note that several
authors in the literature have pointed out and provided analyses of other
cases of two constituents that function as one single lexical unit in several
respects -1 will not assess the analyses; I will only point out some of the
data-.
Pronominal cliticization is the most obvious example of such a
phenomenon. Kayne ( 1975) already analyzed these as Xo items. The degree
of cohesion of pronominal clitics and V elements is very strong, but
arguments have been posited in the literature - cf. especially Kayne
Î1975U19Ô7), Aoun (19Ô5), among others- granting an independent theta
position for clitics, and subsequent movement onto the V head - cf. also
sections 2.1, 3.2 -. According to some authors, the unit cl+V is, thus, not
generated as a unit, but becomes a unit by a syntactic phenomenon
because it involves head movement - see section 3.3.1.1 for further
comments on cliticization-.
Another example of the phenomenon of "non-evident single
constituency" are some sequences of V+preposition - as in sequences of
modal verbs ; i.e. V+P+infinitival -, which seem to function as one single
lexical item as well. G&H (I960) already suggest an analysis in terms of
incorporation of the preposition to the verb; head movement- although
they do not refer directly to Baker's incorporation theory - for the sequence
have * to in English. For it, G&H suggest adjunction of to to have and
they propose a structure as in (l)a.,b.. It must be noted that in their
analysis they assume to to head a PP at S-structure (Da , but to be
interpreted as part of a predicate at LF ( DbJ.e. as a VP- cf. section 2.4, and
3-4- G&H (19ÔÔ); 53-54):
(Da. Johnj I ÍVP has* to/ fpp e/fyp ej leave Mary HI
b. Johnj Hyp has * to/ fype/f ypej leave Mary III
That have-f-to form a syntactic constituent is shown by the
ungrammatically of (2). - G&H (1955); 53 -:
(2 ) *John Ms sot to leave Mary
The same analysis is posited for Italian avere+da , where the
preposition adjoins to the verbal head in the same fashion.
Other verbs requiring a preposition, such as the aspectuals analyzed
in section 2.3.2, are obvious candidates to be considered as non-evident
single consituents: començar * a, tornar + a;empezar + a. volver f a, etc.
Burzio (1986) proposes an analysis for these sequences in Italian - which
he terms prepositional infinitives ( cominciare (a¿ continuare (à), aneare
ía), venire Ca)- in terms of cliticization of each of these prepositions onto
the infinitival verbal head; i.e. the verbal head that follows them. He cites
Rizzi (1982) who gives evidence for such a claim. Evidence such as the
contrast in (3)a. and b., which suggests that the preposition and the
infinitival verb form a syntactic constituent, as they cannot be separated:
(3)a. Alario pensa ene ferse potra partiré
b. * Mario pensa di forse poter partiré
The English sequences of verb plus preposition in phrasal verbs are
also candidates of non-evident single constituents. Stowell (1981) refers to
'Particle Incorporation" and "NP-incorporation" to account for the contrast
in structures like (4)-Stowell (1981); 296- 2, for which he assumed the
structures in (5) - StowelK 1981);301,303- ; where complex X-0 constituents
are created :
(4)a. Kevin turned on the light
b. levin turneé the light on
(5)a. Kevin ÍV'ÍV turned - on 11 the light 111
b. Kevin (y [y [y turned -theiight / - on //
Radford (19ÔÔ) also analyzes such constructions as instances of
word-level adjunctions. An example as (6) is given the structure (7) -
Radfordí 19ÔÔ); 257-:
(6) The weather may turn out rather frosty
(7) v
/ \
V P turn out
As Radford points out: * ... it is traditionally claimed that the [...]
sequence turn out forms a kind of "complex verb". " (p.257)3
As a final observation to this informal introductory section on the
relationship between morphology and syntax - but without implying, that
the here-mentioned analyses are assumed valid - one could mention Rizzi's
claim that two verbs undergoing restructuring become part of one
syntactic consituent - cf. section 2.3.2 -, although the dominating node is
not given an X-0 status - i.e. it is considered a V -. Also worth noting is
Zubizarreta's claim that a sequence of causative plus infinitival complement
is generated as a complex verb in one of its simultaneous parallel
structures - mainly on the basis of the fact that the first verbal element
does not define a stress domain -, and obviously that it forms a semantic
unit with the second verb. One might also mention analyses dating from
Chomsky (1965) where a sequence of different constituents - generated as
separate morphological elements - were assumed to undergo reanalysis and
assigned an X-0 status; typically idioms such as [ taÁ'e advantage of ].
To conclude, the difference between morphologically created lexical
items and syntactically created unite is an issue worth considering. I will
focus on the latter - the former not being the subject matter of study of this
thesis -. Baker's theory of incorporation sheds light on phenomena such as
the ones briefly noted above. Again, as the title of the thesis indicates, I
only intend to attempt an explanation on verbal sequences and not on any
other syntactic units which may have been formed by an identical - or
similar process -. Generally, morphologically created units generated as
one lexical item, remain "opaque" - in a sense to be explained below, cf.
section 3-1.2 - to any_ process whatsoever; on the other hand, two
syntactically independent elements may have the option -cf. 3-3.1- of not
undergoing processes that have the effect of creating a unit, therefore
allowing a margin of "non-cohesiveness". This will give a clue to the
explanation of some apparently problematic interruptions of syntactically
created verbal units - cf 3.3.1.1, and 3.3.1.3. -. It must be noted that, in the
present framework, the borderline between morphology and syntax is not
straightforward, and processes of affixation - in other words, the adjunction
of lexical heads to functional nodes containing inflectional affixes, such as
V-to-I (see section 1.1, 1.2.2) -, although syntactically created, remain as
opaque as "proper" morphologically base-generated units.
3.I.2 The approach
"Si una combinación sintáctica se funde en una unidad léxica, esta nueva unidad es tratada por analogía de la palabra simple y se traslada a ella lo que es posible en relación con la palabra"
(Dietrich (1973:38)
What Dietrich's words express is, to my mind, translated into GB-
barriers terms by Baker in his theory of morphology. In other words, it is
made precise by alluding to concepts like head movement and constraints
that apply to (syntactically created) X-0 elements.
One of the alternative ways in viewing the relationship between
morphology and syntax is the one assumed in Baker(19ôÔ) - basically
following Marantz (1954)- , Ouhalla (19ÔÔ) and others, which regards
morphology as another subtheory in the grammar, on a par with Binding
Theory, Theta Theory, Case Theory, etc. applying to Xo elements at any
level of the grammar, from the lexicon to PF. In this view, Morphology
Theory contains principles that determine the well-formedness of Xo .
Quoting Baker: "As such, "morphology theory" (as we may call it) can be
characterized as the theory of what happens when a complex structure of
the form lz-o X + Y ] is created. In this way, it is parallel to (say) the
binding theory, which is the theory of structures of the form [NPi... NPi' ],
where the subscript is a referential index." (p. 6ô)
Ouhalla (19ÔÔ) claims that, apart from language-specific
morphological principles, there are two universal principles: the Affix
Principle (AP) - (Da.,as in Baker (19Ô5) - and the Head Opacity Condition
(HOC) (l)b.(Ouhalla (19ôô); 15). The Affix Principle is essential as a trigger
for incorporation processes - as will be explained in 3.2 -:
(Ô)a. The morphological subcategorisation frame of affixes
must be satisfied prior to the S-structure level
b. The internal structure of Xo categories is opaque to move-
alpha
Baker (19ÔÔ;73) states (ô)b. as a well-formedness condition on
representations:
(9) * ÍXO ti ]
As Baker points out this may be linked to the Lexicalist Hypothesis.
Selkirk (1962) proposes the Word Autonomy Hypothesis which disallows
rules (i.e. deletion or movement transformations) applying to S-structure
categories to involve categories of W-structure (word-structure). In this
way she rules out the "manipulation of affixes". The intuition in this
proposal is kept in Baker (1986) but it is given another dimension. Selkirk
uses it to disallow the interference of syntax in the domain of morphology,
in other words, to postulate that all morphological processes take place in
the lexicon . Baker uses it to disallow the breaking up of Xo categories
created anywhere in the grammar. Before the creation of a head, move-
alpha may - and in some cases must, as the result of the Affix Principle-
apply to W-structure categories - namely, affixes, which have been
generated in an independent D-structure position. Baker's principle -(9)
above- implies that traces must be exhaustively dominated by an X-0 level
category in the case of head movement.
X-0 categories created in the lexicon are evidently subject to
the same constraints, as implied by the theory of morphology which
regards it as another subtheory.
What the above considerations suggest is that if a sequence of
head categories become one head category by the application of a
syntactic mechanism, the sequence will become a domain of
application of principles which apply to head categories, i.e.
morphology principles.
In the next section I will present the fundamental reasoning
behind the idea that a part of a word - i.e. an affix - which may end
up morphophonologically amalgamated with another part of a word
should be generated in an independent D-structure position; the
principle which ensures it is Baker's Uniformity of Theta Assigning
Hypothesis (UTAH).4 - cf. section 3.2 -.
Another important principle of morphology theory that Baker
posits is the Mirror Principle5 (Baker ( I960); 13):
(10) MIRROR PRINCIPLE:
Morphoiogicai derivations must directly reflect
syntactic derivations (and vice versa)
An example that illustrates this principle , and which Baker uses to
argue in favour of a syntactic analysis of passivization is the following, from
Chichewa, which involves interaction between causativization and
passivization:
(11) KaluJu a-ns-menv- ets-#dw-a kwa ssyanHndl
birimsnkhwi)
hare SP-PAST-beat-CAUS-PASS-ASP to baboons
(by chameleon)
'The hare was caused to be beaten by the baboons (by the
chameleon)"
The causative affix is nearer to the verbal root than the passive affix,
a fact which, according to the Mirror Principle, implies the precedence of
causativization over passivization. Note that - as will be explained below -
the fact that causativization is regarded as a syntactic phenomenon leads to
the postulation of passive also being a syntactic phenomenon. 6
264
3-2 Incorporation (Baker 19ÔÔ)
3.2.O Introduction
Baker(19ôô) posits a mechanism that explains the different
Grammatical Function (GF) changing phenomena found in languages without
the need to postulate different rules for each GF process. He claims that
particular GF changing processes do not exist in language, but that
" "Grammatical function changing" is the side effect of this word movement"
(p.i) ; that is, the consequence of a general mechanism that involves the
combination of two words into one7. By GF changing processes he refers to
constructions where an underlying GF becomes another one at S-st; for
instance, passivization , which involves the change of an object into a
subject - cf. "rule"(14) below -; or causativization, which involves the
introduction of a new subject and the change of a subject into an object - cf.
"rule*'(4) below -. The advantage of such a move is to explain the
relationship between two different sentences with the same meaning
{tà^matíc paraphrases), thus obtaining a unified explanation. Hence he
explains apparently GF changing processes as the incorporation of a V, N or
P into another head category, generally a V - although Note (3D on
passive- ; the consequent structural changes create new government
relationships - with agreement and case consequences - giving rise to what
have been traditionally considered changes in grammatical functions. In GB
GFs are non-primitives , and thus Baker's work is a move towards
eliminating unneccessary notions from the theory.
Baker focuses basically on highly agglutinative languages - Bantu
languages, Eskimo languages, etc - but, as we will see in the following
sections, his proposal seems to be applicable to Romance languages as well -
I will claim this for Catalan and Spanish -. In section 3.2.1 and 3.2.2,1 will
briefly put forward Baker's main ideas giving examples and evidence from
his basic source (agglutinative languages), and in sections 3.3.2 I will show
how he applies these ideas to other languages.^ Section 3-3-1 is an
application of Baker's mechanism to verbal sequences which he does not
consider ; i.e. complex verbs in Romance, basically Catalan and Spanish.
It is of utmost importance to mention that I will only be referring to
a very specific part of Baker's theory; i.e. V- incorporation, because it
seems to me to give a plausible account for verbal sequences in Catalan and
Spanish, basically 9.1 will not refer to the other types of head movement
that he posits.
3.2.I The notion
Incorporation is the term used by Baker for the adjunction of an X-0
element onto another X-0 element. In this section I will sketch how he
makes use of such a process to account for grammatical function changing
phenomena.
Consider the passive example:
U3)a. Alfonso attacked fordi AGENT PATIENT
b. Jordi was attacked (by Alfonso ) PATIENT AGENT
(13) a. and b. express the same relationship between Jordi and
Alfonso, but the surface forms are different. The theta role that each
element bears is the same in both, but the order and morphological shape
266
of the verb differs, plus the fact that the AGENT in U3)b. is introduced by
a preposition. The descriptive characterization is given by Baker as follows:
( 14) Passive: subject --J oblique (or null) ; object —> subject
But (14) is a descriptive statement; it adds nothing theoretically
interesting to the grammar.
Consider an instance of causative in Chichewa (Bantu):
( 15) a. Mtsuko u - na -gw - a
waterpot SP-PAST -fall -ASP ( 10)a.p. 10-11
'The waterpot fell"
b. Mtsikana a -na - u -gw - ets - a mtsuko
girl SP-PAST-OP-fall-CAUS-ASP waterpot (lô)b.
"The girl made the waterpot fall"
Note: where OP = object agreement prefix
SP= subject agreement prefix
This is an instance of one of the different possible causativization
processes in languages and its descriptive statement is :
(16) null —> subject; subject-->object
(which accounts for the agreement relationships indicated)
( 15)b. may be paraphrased in Chichewa as in ( 17) ,
(17) Mtsikana ana - chit-its-a kutJ mtsuko u-gw-e
girl AGR - do-MAKE -ASP that waterpot AGR-fall-ASP
"The girl made the waterpot fall" (2)a. p. 148
Just like (13)a. and b., 05)b. and (17) also express the same
"meaning" relationships between NPs. Baker's way of characterizing the two
corresponding structures is by saying they are thematic paraphrases. This
is a clue to their underlying structure. Namely, Baker,formulates a principle
which will ensure that, thematic paraphrases have the same D-structure,
but, for reasons to be explained, surface differently.
(10) The Uniformity of Theta Assignment Hvpothesis:(UTAH)
Identical thematic relationships between items are
represented by identical structural relationships between
those items at the level of D-structure (30) p.46
The UTAH implies that the passive counterpart of ( 13)a. has the same
D-structure as (13)a... the active form: the thematic relationship between
attacl'ecf and prdi is identical in both, so it must be represented in an
identical structural relationship. For the Chichewa examples, the UTAH
applies to part, of & word. Quoting Baker (p.49) : "Generally, whenever part
of a word shows signs of assigning or receiving a thematic role in the same
way that morphologically independent constituents do, the UTAH will imply
that that part of the word appears in an independent position at D-
structure"
The relevant structures are :
(lcj)Passive: S 10
A NP VP
A V NP
sttsct prâi
The theta-role PATIENT is assigned canonically. The passive account
in GB is that the NP-obi must move to subiect Dosition in order to get case J J A KJ
since the participle has been devoided of the ability to assign case by the
morphological process deriving the participle - cf, also alternative analyses
such as Jaeggli ( 19ô6) where the participle affix absorbs theta-role -. Baker
explains passive differently: he accounts for it in terms of the incorporation
of the V into I , thus in this sense, claiming that it is a syntactic process,
and not only a morphological one. He accounts for the status of the implicit
argument by claiming that I is an argument and must be assigned a theta-
role . In this way , passive in English is regarded as an instance of
incorporation, although a different sort of incorporation - as it involves a
head movement into a non-lexical head, I - cf. Note (32) on Baker's account
of passive.
(20) Causative:
NP VP 0TJ / \ ^
V
NP VP
pût
V
fall
As (15b) shows, fall is not in its D-structure position at S-structure,
movement takes place and gw - fall- morphologically combines with its -
mate -. Incorporation, thus, implies movement; it is an instance of move-
alpha. Movement, though, is not triggered by the usual case or by +wh
COMP features. Here, the trigger of movement is the morphological
subcategorization of the elements that appear separated in D-structure:
they are affixes and must be bound - cf. the Affix Filter in 3.1 -. Baker
follows Lieber (I960) and Williams (1961) in saying that affixes are
specified for all the same types of features as independent words, and he
goes a step further in assuming that elements which surface as affixes may
head phrases and assign theta roles just like independent words do, at D-
structure. The difference between words and affixes is, then, "that affixes
must attach to a word - clearly a morphological requirement. If an item is
specified as being an affix, but is generated in an independent position at
D-structure in accordance with the UTAH, that item will have to undergo
/N
270
X-0 movement to adjoin to some other X-0, failure to do so will result in a
structure which results in the violation of a principle of morphology." (p.72)
3.2.2 The process: V-Incorporation (VI)
As the passage quoted above states, affixes - X-0 categories at D-
structure - must move to adjoin to another X-0 . Move-alpha, thus will not
imply the movement of an XP but rather of a head of a phrase in
incorporation structures.
This type of movement is not new: Chomsky (19ô6b) - cf. section
1.2.2 - , Koopman (19Ô4), Torrego (1954), Pollock (I960) among many
others have assumed that V is subject to move-alpha under certain
circumstances. As mentioned, Baker applies X-0 movement to V, N and P in
order to account for GF changing phenomena of different types.
Reviewing certain conclusions reached in outlining the barriers
framework - section 1.2.2 -, there are several important facts about X-0
movement. Firstly, a head may only move to another head position and
adjoin to it; it may not move to an XP position, a configuration which would
presumably violate some version of the Structure Preserving Hypothesis
(Emonds 1976) -. By the Projection Principle, an X-0 category must leave a
trace; if lexical properties of items are to be kept throughout the derivation,
an element which assigns a theta-role to a certain position at D-structure
must leave a trace, just like an element which has received a theta-role in a
certain position at D-structure. The structural relationship between the two
positions created by movement must be that of government, a consequence
of the fact that all traces are subject to the ECP - the syntactic aspect of
incorporation -. Since X-0 categories cannot be lexically governed - by
virtue of the fact that theta-marking takes place only between sisters - cf.
271
section 1.2.2 where theta-marking is shown not to percolate to the head of
a theta-marked Xmax - antecedent government must be satisfied.
It must be noted that Baker somewhat modifies the notion oí
barrierhood in that instead of BC and L-marking, he introduces the notion
of s#JectJoj2 as follows: (p.56-57)
(21) A selects B if and only if:
(i) Á assigns a theta role to B
(ii) A is of a category C and B is its IP
(iii)A is of a category I and B is its VP
(50)
He also accounts for minimality barriers (22) (ii) :
(22) Let D be the smallest maximal projection containing A.
Then C is a BARRIER between A and B if and only if C is a
maximal projection that contains B and excludes A, and
either:
(i) C is not selected, or
(ii) the head of C is distinct from the head of D and
selects some WP equal to or containing B. 1 1
(49)
These modifications together with Baker's indexing system have
the same results - as far as V-incorporation in the cases considered
are concerned -than the assumptions in Chomsky (1906b) : an X-0
trace must be antecedent governed. To summarize Baker's indexing
system, there are three types of indices: identification indexing (the
result of movement); theta-indexing ( the result of theta-assignment);
and Case-indexing ( the result of Case-assignment).
Baker states the ECP in terms of a requirement of either theta-
indexing or identification indexing. Theta-indexing is never available
to traces left by incorporation because theta-marking can only take
place between sister nodes, as already noted. Hence, by assumption,
the theta-index that the XP dominating the trace of the incorporated
item bears never percolates down to it. Thus, the only possible way
for an X-0 trace to be properly governed is for its antecedent to
govern it. Baker states the ECP as follows for X-0 traces:
(23) An X-0 must govern its trace
Government requires two things: that c-command hold between
the antecedent and the trace, and that there be no barriers between
the two elements. Both are satisfied in proper incorporation structures,
as will be exemplified below.
With the joining of two X-0 level categories , incorporation
creates a syntactic node which does not comform to the X'-schema in
the sense that it is not an X' category - it has no complement position-,
and it is not a proper X-0 category - it contains two different heads.
The special status of this category is expressed by the following
statement as regards its interpretation:
(24) The indices of the parts of an X-0 category count as
the indices of the X-0 category itself
In the structure:
273
(25) YP
Y* XP
Xi Y ti ZP
Y* c-commands the trace and is a proper governor ior i t because
it bears tüe same identification index, by vir tue of the fact tha t i t
contains the antecedent of the trace.
The definition of barrier is somewhat modified by Baker , as
stated above. In other terms, for Chomsky, the categories tha t can be
barriers between two elements, A and B, are relative to the B element
- cf 1.2.2 -. Baker makes the notion of barrier relative to both, the
governor and the governee:
(26) The maximal projection C is a government barrier between
A and B if and only if C contains B, C does not contain A,
and C is not theta-indexed ( with A ) 1 2
In (25), repeated here with the corresponding theta-indices:
(27) YP
Xi Yj ti ZP
274
XP is the only potential barrier between X and t; it does not
contain Y* (=A) and it contains t ( = B), but it is not a barrier because it
is theta-indexed with A (=Y*) - Y assigns a theta-role to XP, and thus
shares the same theta-index with it - and by (24) Y* also shares that
same index.
Another condition that must be satisfied by all instances of
movement is Subjacency. Subjacency with regard to incorporation will
always be redundant since the crossing of just one barrier will block
government, resulting in an ECP violation, and, thus, the trace will not
be licensed. If the X-0 moves to a Y-0 that is not theta-indexed with
the XP headed by the X, the XP will be a barrier to government since it
will not satisfy the theta-requirement.
Another propsal by Baker is the Government Transparency
Corollary (GTC):
(2ô) A lexical category which has an item incorporated to it
governs everything which the incorporated item governed
in its original position
This follows from the indexing relations mentioned above , as
illustrated in the following struture:
(29) YP
Xi Yj ti ZPi
The notion oí transparency implies basically that neither the XP
nor the ZP are barriers to government when the head of the XP is
incorporated. This will have a series of consequences in Grammatical
Changing processes in most of the languages that Baker studies: a
complex verb will govern more elements once it has been incorporated
and, thus, more agreement possibilities arise giving rise to apparently
GF changing process. The GTC has no consequences for the kind of
complex verbs for which I will propose incoporation - cf. 3-3-1 -.
The ECP, thus, is satisfied in a head-to-head movement
structure such as (3D) - the Chichewa causative, to be considered again
below - if the movement of the lower verb follows the Ifeaä
Movement Constraint (3D - cf. also section 1.2.2 -:
V V
fallj mate
276
(3D HEAD MOVEMENT CONSTRAINT(HMC)
An X-0 may only move into the Y-0 which properly governs
it. (43)
As shown by Baker, the HMC is derived from the ECP since it
implies that traces of x-0 must be properly governed. This is the case
in incorporation structures.
Biclausal structures
There are two possible situations where movement of a head
may take place:
1. if the head is governed by another head in its D-structure
position it may incorporate into the other head following the HMC
straightforwardly
2.if the head is not in a governed position, it must move prior to
incorporation in order to reach a position from which it is governed by
the X-0 category into which it incorporates. This is the case of biclausal
structures, as we will see below.
Consider first Baker's account of the following incorporation
structures:
(32 )a. [yp Xi + Y [ xpti ZP]] (where XP is selected by Y)
b* [ YP Xi + Y [ XP ti ZP ]] (where XP is not selected by Y)
c.*lYPZi + Y[xpX Izpti]]] (40)
(32)b.and c. are both ruled out by Baker's notion of barrierhood:
b. implies an ECP violation because the XP is not selected, and c.
implies a violation because there is a head, distinct from Y which
selects ZP containing the trace of Z; in other words, SP is a minimality
barrier in c. In a structure such as this, incorporation would only be
allowed if Z would have moved to a position from which Y could
govern it prior to incorporating - X -. Direct movement is not allowed:
it violates the HMC, or, alternatively, the t violates the ECP.
(32a.) is an instance of situation 1.; it is either a simple clausal
structure like (30) above, where S is selected by V, and there are no
other maximal projections intervening, or the movement of the head
of any selected complement - for instance N incorporation to V -.
Now consider the structure assumed for a complex configuration,
in the case of the Chichewa causative:
(33) IP
NP
gití
Y
VP
V CP
mak& \
C"
IP
NP I
wsterpot.
I VP
V
fan
Direct verb incorporation - henceforth VI- is not possible: there
are two head positions which could create minimality barriers -
namely C and I- and cause an ECP violation. It is relevant to point out
the notion of distinctness that Baker defines - cf. also (22)(ii) - in
order to fit his barrierhood notion with the incorporation structures :
(34) X is distinct from Y only if no part of Y is a member of a
(movement) chain containing X (64)
279
This notion is relevant because if the head in question contains a
trace of the incorporated element, then the maximal projection
dominating it will no longer be a minimality barrier; its head will not
be distinct from the potential governor (as long as the governor
contains the antecedent): An instance of this :
05? Vt
y* sp
Xi Y X ZP ! I ti Z
In this structure, XP cannot count as a barrier between Y* (a
potential governor) and ZP because XP contains the trace of X; it is not
distinct from Y*.
Considering structure (33) above, the lower verb must.
incorporate into the higher verb, due to its morphological
requirements. There are two ways for the lower verb to be able to
move into the higher verb:
a. from the Spec-CP position (if the whole VP moves)
b. from the C position (if only V moves)
Baker claims that both possibilities are attested, and give rise to
different causative structures. Baker analyzes complex predicate
structures in Romance as instances of a special kind of
incorporporation - abstract mccrpcratJojx which I will summarize in
section 3.3.2 - as instances of the a. possibility. I will, thus, leave the
b. possibility unsummarized as it is assumed for languages which are
not the subject of this thesis. Nevertheless, direct V-to-V movement
will be claimed for complex verb sequences -cf. section 3.3.1 -.
Note that in structures where a CP node is assumed to intervene
between two verbs to be incorporated, VI involves the incorporation
of a head within a (selected) complement. The embedded internal
movement of the V to reach a position from which to incorporate is
not directly related to the status of the CP, but the next step, proper
VI, is directly related because CP must be selected in order not to
count as a barrier between V* and the trace of the incorporated V.
3-3 Incorporation and verbal sequences
3-3-0 Introduction
Section 3.3.1 represents an attempt to apply the mechanism of
incorporation to complex verb sequences in Catalan and Spanish. A
fundamental difference regarding the languages that Baker studies in more
depth and the two Romance languages for which I will propose
incorporation is the issue of the trigger of incorporation. As mentioned,
languages like Chichewa- in the constructions where incorporation is
assumed to take place- the incorporated items constitute a single word, a
morphologically complex item that surfaces as a unit. The Affix Principle
(AP) requires that affixes which are generated in independent D-structure
positions move to positions where they may attach to an affix-bearing
element - there may be cases in which they may attract a host to move to
their position -. This morphological justification for movement is obviously
absent in both Catalan and Spanish. The verbal sequences do not form one
word; they surface as two separate morphological units - crucially, that the
first item bear independent inflectional morphology is evidence for this.
The reasons that bring one to postulate incorporation for complex verbs in
Catalan and Spanish when there is no direct morphological evidence stem
basically from the fact that they function as a single lexical unit. Apart from
the tests which will be alluded to in ¿ection 3.3.1.1, an important piece of
evidence is the fact that the first member of the complex verb - &a/ra -
does not define a stress domain, and it has a clitic status - but see also
3-3-1-1- The complex verbs introduced in section 2.2.2 are , thus, analyzed
in 3.3.1 as instances of head movement in the syntax. I must emphasize
that although cliticization is posited as subsequent to head-movement, the
262
fact that cliticizatíon represents the interface not only between morphology
and syntax but also phonology overreaches by far the scope of this thesis,
so no detailed analysis may be provided here.
Section 3.3.2 presents the type of incorporation that Baker postulates
for Romance complex predicates , abstract mcorporatJon ( or r$anajysjs).
The basic idea is that in the sequences for which this type of incorporation
is assumed the two verbal elements are semantically linked, they constitute
a complex predicate - cf. section 2.3 - but not a complex word; according to
Baker, their "amalgamation" must be at a different level: " Fare is not an
incorporator, but a "reanalyzer" ( an LF affix?) and must enter into the
Reanalysis relationship with another verb at LF. This may be a semantic
property of the verb, to the effect that it forms "complex semantic
predicates", since it is generally the same kinds of verbs which have such
properties from language to language" (p. 203). I will refer only to the
essentials of Baker's abstract mcorporatJon proposal, and propose some
arguments against it in section 3-4 , suggesting that G&HÎ19ÔÔ)' s proposai
gives the theory sufficient scope to account for complex predicate
behaviour without having to postulate LF-movement.
Note that there is another way of characterising the verbal sequences
in question - sections 3.3.1 and 3.3.2; namely, as verbal clusters or verbal
chains - cf. Bordelois & Evers (1990) - . This would lead to say that those
sequences in 3.3.1 - complex verbs - are verbal clusters, although the fact
that they are analyzed as head-movement implies that they are also part of
a chain, which includes empty Xo categories subject to the ECP; and that
those sequences in 3.3.2 - complex predicates - imply non-movement
chains - in line with G&H (19ôÔ)'s proposal; cf. 2.4 and 3.4 -. *3
3.3-1 Incorporation and complex verbs
3 3 1-1 Complex verbs as x£ units
Reconsider some or the tests presented in section 2.1 as tests for
auxiliaries versus main verbs:
Catalan;
1.- Interruption by lexical items (adverbs and parentheticals)
(Test 10)
(36)a. He vlst la Teresa al bar fa una estona
b* He, fa una estona. vlstla Teresa al bar
c* He casualment vist Ja Teresa al bar
(37)a. Be veritat hem cregut que t'agradarla
b* Hem, de veritat, cregut que tagradarla
c* Hem sincerament cregut que t'agradaria
(3ô)a. Va veure la Teresa al bar
b * Va.ahir al vespre, veure la Teresa al bar
c* Va casualment veure la Teresa al bar *
(39)a. Vam saber que tenia molts diners
b. * Vam sempre saber que tenia molts diners
2. Placement of the negative particle (or Tensed negative VP)
( Test 3 )
(40)a. La Maria no va venir
b. *LaMaria va no venir
(4 l)a. En Terendnona entrevistat en Lfiu's
b. * £0 Terendli3j20£nïz£EM2Len Liais
\ VQ movement ( or Tensed verb preceding subject):
( Test 4 )
(42) a. Ha trucat en Joan?
b. * Ha en Joan trucat?
(43) a. Va veure el partit el Llius?
b. * Ya en Lluís veure el partit?
4. Preposing and postposing
( Test 11)
(44)(FromH&R 19Ô4)
a. Va parlar de le seves velles amistatsparisenques durant l'estiu
b. * Va durant l'estiu parlar de les seves velles amistats
parisenques
(45)a. Ban dit que farien els deures i els nan fet
b * Han dit que farien els deures i fet elsnan
(46)a. Creia que veuria l'eclipsi i el va veure
b. * Creia que veuria l'eclipsi i veure'l va
The examples above show a basic characteristic of Catalan verbal
sequences : the sequences made up of "haver"+ participle and "va"+
infinitive - i.e the sequences which I have called complex verbs - may not
be interrupted by other constituents nor may be subject to move-alpha.
The verbal sequences labelled complex verbs form an indivisible unit.
Spanish:
The following examples seem to indicate that Spanish has the same
restrictions as Catalan as regards the interruption and breaking up of
verbal sequences which constitute complex verbs.14
1. Interruption bv lexical items ( adverbs and parentheticals ):
(Test 10)
(47)a. He visto a Teresa en eJharnace un rato/casualmente
b. * J&_. ¿tace un raie, visto a Teresa ex eJ A?/-
casualmente
(4ô)a. De verdad /sinceramente nemes creido que te gustaría
b. * Hernes, de verdad, creído que te gustar/a
sinceramente
2. Placement of the negative particle ( or Tensed negative VP)
(Test 3)
(49) a. luis ne A? JJamado
b. * Luis nano JJamado
\ V movement ( or Tensed verb preceding subject):
( Test 4 )
(50)a. Ha JJamado ¿uis ?
b. *Ha Litis llamado ?
(50a. Habrán llegado los invitados ?
b. * fíabrÉR los invitados llegado ?
4. Proposing and postposing:
(Test 11)
(52) a. Lilis dijo quo vendría y ha venido
b. * Luis dijo quo vendría y venido ha
(53)a. Los alumnos dijeron que estudiarían vhan estudiado
b. *Los alumnos dijeron que estudiarían y estudiado han
Now, if we consider these non-evident single constituents - (36)-
(53X- and assume they are X-0 constituents- pending a syntactic derivation
in section 3.3.1.3 -, it seems that the HOC - as defined in section 3-1 - may
be made responsible for their behaviour: the V movement - i.e Vo
movement and preposing/postposing - examples show that apart of the
sequence cannot be moved, under any circumstances -i.e. that there can be
no V trace of part of the sequence -. Nevertheless, apart from this fact, we
could also appeal to the HOC to account for the impossibility of interruption
of the sequence. One could modify the HOC by making it general enough as
to ban any type of separation of a category once it becomes a proper l9
category, and one could attempt a defintion of "proper" YS> category by
implying morphophonological amalgamation. - in the case of incorporated
sequences complex verb sequences which have undergone head-movement
and subsequent cliticization - see below " A note on cliticization" -
.Generalizing the notion of "opaqueness" to include other ways in which a
head becomes "frozen/opaque"(Ouhalla (1908,16), the HOC may be
reformulated in the following terms:
(54) Xo categories are opaque
If a set of elements is dominated by an Xo node these elements are
not accessible to movement, plus they may not be interrupted by other
constituents.
A note on cliticization
As mentioned, the facts presented so far obviously lead one to the
issue of ctfttasatfaa Clitic climbing was already alluded to as a test to show
the cohesiveness of complex verbs and complex predicates - cf. 2.1, and 2.3
-, but an important distinction between "clitic climbing" -as in Kayne( 1975),
Burzio(19ô6) among so many others - and "cliticization" m complex verb
sequences must be made as well as some qualifications added to this issue.
It is well-known that clitics are a not easily defined set of elements.
The criteria that work for clitics in one language may fail to work for those
elements considered clitics in a different language - cf. especially Spencer
(forthcoming) for definitions and comparison of clitic systems -. Two
definitions will do to see that capturing the notion of cMtJc is not a simple
matter:
"It is largely agreed that genuine clitics are words which happen to be phonologically dependent on a host. Thus, they are elements which have the syntactic properties of words, but the phonological properties of affixes." (p.7ô) and from the same source: ".. we will see yet more evidence of schizophrenia on the part
2ÔÔ
of clitics, behaving now as fully-fledged words, now as ¡hound morpnemeS (p. 16)
(Spencer (forthcoming))
"Solem designar amb el terme "clitic" aquests elements morfològics que ocupen l'espai entre el mot i el morfema. En aquest espai aquelles propietats que diferencien el mot del morfema ( caràcter lliure, certa llibertat d'ordre, fenòmens fonologies típics, inexistència d'al·lomorfia ) esbarregen..."
(Mascaró (19Ô5)p-12 3)
(italics mine)
From these definitions one may expect to find difficulties in trying to
account for the behaviour of sequences of elements that contain a clitic, as
is the case with complex verbs. As is pointed out in both definitions, clitics
have some properties of words, and some of affixes. Clitics are not like
words, for instance, in the fact that they do not ususally bear stress, and
their clustering is not hierarchical like that of syntactic constituents. "... we
are actually dealing with an unusual form of free-floating affixation , and
that as far as the syntax is concerned clitics aren't words at all " ( Spencer
(forthcoming) p. 14). But clitics are not affixes either: they may attach to
different types of elements in many languages - for instance, Serbo-Croat -,
and they may move to different positions, a rather word-like property.
In his comparison of different clitic systems - Serbo-Croat,
Macedonian, and Portuguese - Spencer clearly shows that even
characteristics which seem general enough of clitics are not universal. As he
points out, following Klavans (19ô2)'s observation, the fact that clitics do
not bear stress is not a universal characteristic of clitics. Another important
apparently general characteristic of clitics - that of being tightly knit with
their host- is also not universal. These two observations - plus others - will
allow us to explain some otherwise problematic examples in Catalan and
Spanish; i.e. sequences where other clitics interrupt the sequence of a
complex verb, or sequences where the first element in a complex verb
sequence bears stress.
The formation of the future tense in many Romance languages, is a
consequence of a content word becoming an affix. Catalan and Spanish are
examples of this - see section 3.3.1.3 -. Portuguese seems to have followed
the same development with the only difference that the future suffixes
may be separated from their host verb by a pronominal clitic as in the
following examples from Spencer: ter&fJeyadoi = I shall have raised ) / / Ä?
- Jo - OJ levado (= I shall have raised it ). This is in clear constrast with
Catalan and Spanish where the future suffixes are not separated from their
host not even in the compound tenses. ! 5
Classifications of clitics include not only pronominal clitics but also
auxiliary forms, determiners, conjunctions, and prepositions. As the subject
of this thesis are verbal sequences, I will only refer to the clitic status of
auxiliaries in complex tenses - except for the observations in Chapter 2 on
Clitic Placement/Climbing as a test for cohesion between two verbs -.
Nevertheless, a crucial difference with pronominal clitics and auxiliary
clitics must be noted: due to the argument status of the former, there has
been intense debate *& claiming either movement from a base-generated
original position or base-generation on their host, a debate which has not
taken place for auxiliary clitics.
To conclude this note on clitics - to be further discussed as applied to
complex tenses in Catalan and Spanish in 33-13- - I will point out an
observation by Mascaró (1966) which, I believe may be traslated into
incorporation - plus subsequent cliticization - . I am referring to the
comparison that he makes between compounds and "larger words" - "small
words" plus clitics -: he distinguishes between the lexical character of the
former-compounds- and the syntactic character of the latter-larger
words"-, paving the way for a syntactic characterization of larger words",
as is the case with complex tenses.
3.3.1.2 The structure of complex verbs
In this section I will put forward the structure that has been
independently posited in many works - cf. especially Zagona (19ôô), among
others - and which will provide the possibility of postulating direct
incorporation of V to V in complex verb sequences. - 3.3.1.3 - 1 7
This section is in a sense the attempt to show that even if the UTAH -
cf. section 3.2 - of Baker is not involved in the structure of complex verbs,
there are still reasons to believe that the two Vs are not generated as a unit
but become a unit in the syntax; each is granted an independent head
position. We will assume for the time being - and reconsider the issue in
Chapter 4 -, that each V heads a VP.
In section 2.2.2 I briefly reviewed a few proposals, some of which
were within earlier frameworks - ASWÜ979), Emonds (1976) -.
Nevertheless, each still granted each "auxiliary" a separate V node - not yet
a "head" in terms of X'-theory. The structures are reapeated here:
Emonds (1976)
V V
lavés
V V A
ASWC1979)
(5«
(have)
V
Zagona(19ÔÔ) represents one of the reformulations of these proposals
in X"-theory terms : each auxiliary heads a VP with full phrasal structure; it
has all possible projections - as explained in section 2.2.2 -. The structure is
repeated here:
(57) V
spec V (XP)
where XP is a possible position for a modifying adverb, and V" the
complement, of an auxiliary.
Her theory, moreover, puts forward the licensing of Vs
subcategorizing for VP, vritb her specific temporal-role marking proposal.
Furthermore, the parameter she proposes to distinguish English and
Spanish does not rely on a different VP phrasal structure for auxiliaries ,
but rather on different possibilities for temporal assignment.
The notion of complex predicate in Manzini (19Ô7) also involves
more than one VP: a "verb-to-verb" selection allows for the properties of
the non-auxiliary head to be "dominant" as far as the selection of the
subject is concerned.
ManzinidÇjô?)1^
(SÄ) VP
/ \ NP VP
/ \ ï v :
V VP
I V
V VP
In G&H (I960)- cf. section 2.4 - it was shown that auxiliaries T-mark
VPs, so subsequent VPs are also permitted and LF licensed by the FDC. An
example (£35);49) of a complex verb in their paper is repeated here - cf.
also 3.4. -:
(59) John íj'Iívp hasfvp laughed///
Another basic source which validates consecutive VPs ; i.e. the
subcategorization of VP by V is Chomsky ( 1956b) 1% where such a proposal
is not argued for or explicitly claimed. Nevertheless, a consecutive VP
structure is given for passive - cf. also 1.2.2 - . The structure of passive
after V-raising is the following, where @=i=j by Spec-head agreement:
((171;p.76)
(to) Jiaaoj fate -JHyp- tj fyp killed tjI
331-3 Syntactic incorporation
As made explicit in the previous section , the structure assumed for
complex verbs in Catalan and Spanish is:
(61) VP
V
/ \
Vl VP
\ V2 . . .
partidple/infiniiive
In such a configuration, the relevant structural relationships hold of
the two V-elements in question so that move-X0 may take place: V2 moves
to adjoin to Vi and such a process results in the creation of a new V-
category, V*.
(62) VP
A Vi* VP
A I Vi V2 V2
«once matme v? position is aireaay governea by vi prior io movement.
The HMC is satisfied, and consequently the trace of V2 does not violate the
ECP. As explained in 3-2, the V* counts as an antecedent governor for the
V2 trace; they share indices20 and, thus, antecedent government holds.
Note that V i to V2 incorporation is ruled out:
Vi V2
This is a case of downward movement, where the trace c-commands
its antecedent, and, the ECP is violated, government of the trace by V* is
not possible here as there is an intervening maximal projection which,
although selected, blocks c-command.
in structure (ôz), once the v-2 nas moved up to v-l, there is nothing
• to prevent a cliticization of the first verbal element onto the second. This
gives the expected results, i.e. the impossibility of separating them, which
parallels with the behaviour of all other clitics in both Spanish and Catalan:
(64)a. sempre Ja veJg
b. Ja veig sempre
c. * Ja sempre veig
{h^)z..sJempre Ja veo
b. Ja ¥$o siempre
c. * Ja siempre veo
Therefore, once incorporation has taken place, cliticization follows,
although, a qualification must be added at this point, which will be taken
up again in the next section. The fact that clitics tend to be monosyllabic
and not bear stress - although, as pointed out in section 3.3.1.1. in "A note
on cliticization", this is not a universal - implies that cliticization will be
ensured when the first verb in the sequence is monosyllabic, but non-
cliticization may give a certain degree of non-cohesiveness in non-
monosyllabic forms - cf. Suñer (19ÔÔ), section 3-3-1-4 -.
If we assume that the process of incorporation followed by
cliticization leads to a "frozen/opaque" constituent for the HOC - cf.3.1 -,
the existence of an element which interrupts the complex verb sequence in
Catalan - jaas - may be due to its own status as a clitic - in these
constructions21 - preceding cliticization of the first verbal element:
(66) Noüepas vist Ja nena
Note that, as was already observed, it may be expected of a unit that
becomes an Xo constituent in the syntax to be less tightly knit than a
lexical unit created in the lexicon. Although, as implied by Morphology
Theory they are both subject to the same constraints. Moreover, as pointed
out in "A note on cliticization" in section 3.3.1.1, there are even cases - as
in Portuguese future tense - where a pronominal clitic may intervene
between an affix - which has historically developed into one from a
previous content word - and its host verb. Phenomena such as these may
be expected in an area as murky as is cliticization. 2 2
As has been noted in section 3.3.O, the trigger for postulating
incorporation in languages like Chichewa is the Affix Principle. The V, N, or
P elements that undergo incorporation in these languages are generated in
a different D-structure position by virtue of the UTAH - cf. 3-2 -, but they
are affixes. As such they need a host, and to find one, they must move. In
complex verb sequences for which I am proposing incorporation there is no
such morphological justification to allude to; there is no morphological
trigger. The assumption that I will be making here is that verbs that
subcategorize for VPs trigger incorporation of the participle - or infinitive,
in the case of the pt?+infinitive sequence in Catalan - ; i.e. they attract the
main verb onto them. Incorporation in such cases is obligatory - with
possible subsequent cliticization -. 23
Why aren't the two V-elements phonologically amalgamated? The
answer to this question is beyond the scope of this thesis, but one could
attempt an approximative answer by alluding to a language-specific
morphological characteristic of Catalan and Spanish - and all Romance
languages, and also English- : the lack of inflectional prefixes . This might be
the reason why , although HAVER/HABER has lost its lexical (= "possess" )
meaning in modern Catalan and Spanish, they have still retained a lexical -
V - and non-affixal status.
There is a phenomenon in the history of both Spanish and Catalan
that may shed some light on the behaviour oí complex verbs: the formation
of the Future tense - cf. also section 3.3-1.1. for Portuguese - . The present
day Future tense(b) is the fusion of two forms; an infinitival form and a
form of "haver/haber"(a) :
(67)(a) amar + he = (b) amaré
amar + has amarás
fer + he = (b)faré
fer + has faràs
We could explain the development of these forms by positing a
process of incorporation. The difference with the complex verb sequences is
that one could allude to a possible formation of the future forms by a left
incorporation of V i to V2 , as in (60):
(66) VPi
/ \ Vi VP2
/ \ \ V2 V i V
f&r i¡e J
V2
t
As is usually assumed, adjunction is allowed both to the right and to
the left, but the morphophonological result is different; in the future case,
v
29Ô
the fact that inflectional suffixes do exist in the language may have paved
the way for the amalgamation. It must be stressed that this process may be
the explanation of a diachronic phenomenon; it is obvious that present-day
future inflection is an affix and not an independent lexical head - as I am
assuming for V1 in complex verb sequences -.
There is another problem that arises if we assume that the complex
verb sequence becomes an I o unit in the syntax: the fact that the first
element bears tense and agreement; i.e. why are the (b) examples not the
result of the union of the two verbal elements?:
(69)a. Has fet
b. *ha fûts
(70)a. has techo
ï>.*MJ3$CÙOS
Here, again we allude to the fact that morphologically created units
and syntactically created units are not completely alike, although they are
both subject to the HOC. There seems to be only one way to account for this
fact: assuming that the syntactic process is a "special" kind of adjunction;
proper adjunction would lead to the T and AGR morphemes to be borne by
V2 -as would be expected of the structure (30) -. but since the participle
has already undergone morphological processes - as postulated in several
recent studies; cf. Belletti(1990), Drijkoningen (19Ô9X among others.- no
more affixes can attach to it. Note that there is an independent ban on
gerunds and participles to bear any person or number affixes. The
following structure includes inflectional nodes and expresses the
subsequent adjunctions that have applied: V2-to-V 1 -to-T-to-AGR24 :
299
(71) AGRP
V V
V
t
A comparative fact to point out in favour of the clitic status of the
auxiliaries in question, despite their bearing independent inflections is that
Serbo-Croat auxiliary clitics have this same property, they inflect for tense
and for person. - cf. Spencer (forthcoming) -.
3.3.1.4 Remarks on language variation
In this section some observations are made regarding the different
behaviour of equivalent complex verb structures basically in Catalan and
Spanish, but also Italian and a note on English and French is made.
Spanish
There are certain differences between Catalan and Spanish that call
for an explanation. Suñer (I960) reports a contrast found in Spanish which
does not seem to arise in Catalan. Namely, the contrast between one-
syllable forms of the verb "haber", and longer forms.
Suñer gives the following examples to argue against the "inseparable
unit" hypothesis:
(72 ) Lo hubiera usted oído en Ja mesa (4)a.
(73) ¿ Como se espJJca çue desde entonces ses no naya usted dado
paso aJgiwo? (4)b.
(74) ¿ Como hubieras tu quedado? (4)c.
(75) .... ycuando yo hubiera ya sido mayor... (4)d.
(76) J>;' despues i?ue había mucho corrido, ¿no? (4)e.
(77) Habían eJJospagado ya anticipadamente (5)a.
(76) Platero me habia ya saludado con un rebuzno (5)b.
(79) ße ahí que hubiese, durante meses, visitado Ja casa sin toparse
con Marià (5)c.
(00) ¿çue habrá ef gobernador aprobado Ja semana pasada? (6)a.
(Ô1 ) cí4 quien hubiera el vecino agradecido Ja atención si... ?(b)b.
(Ô 2 ) Se había casi convertido a Ja reJigion de su novia cuando... (6)c.
(03) £sto habría indudabJemente aceJerado eJproceso (6)d. «
The above - also from Sufîer(19Ô7);6ô3 - contrast with:
(04) * ¿Jía Pepe terminado eJJJbro? (Strozer 1976)
(05)* ¿çuéha Juan Jeído? (Zagona 19Ô2 )
(Ô6) * ¿Çuéha Ja gente organizado? (Torrego 19Ô4)
The examples cited show that there are instances of the verbal
complex "haber + participle** which allow interruption. It must be noted that
the dialects that Suñer considers do not coincide with the dialect of many
Spanish speaking native speakers of the Iberian Peninsula. For many
speakers, examples (72)-(ô3) are not ungrammatical but are not wholly
acceptable either.
Suñer explains the difference by alluding to a cliticization of the
monosyllabic form of the verb "haber*" which needs a host and cannot stand
on its own. Evidence for this comes from the fact that the only monosyllabic
form of the existential "haber" (there-be) ha does not exist as such; its only
form consists of ha+a locative clitic "y" which has not survived in modern
Spanish:
(07) Hay muchagente en este Jugar
(Ôô) Bab/a mucha gente en ese Jugar
(07) and (ÔÔ) illustrate the difference: only the monosyllabic ha
needs the clitic "y", the past form ha&fa does not. Suner, thus explains the
contrast by saying that when one of the monosyllabic forms of the verb
"haber" occurs - lsg he, 2sg has; 3sg ha, 3pl han -, it always cliticizes to the
form that follows it.
Note that - irrespective of the validity of Suñer's analysis - the fact
that monosyllabic forms need a host points towards the possibility of
incorporation applying in this sequence - as the mechanism to license
cliticization -. But, note that, universally, clitics are not always
monosyllabic; as for instance the bisyllabic prepositions of Catalan and
Spanish - cf. also Spencer (forthcoming), Klavans (19Ô2) -.
Catalan
Consider the Catalan translations to some of the above sentences:
(72') * Si l'haguës vostè sentit a la taula
(74') * Com hauries tu quedat?
(76') * Jöf després çue havies molt corregut.
(77") *Havien ells pagat, ja anticipadament
(79") * I> aquí que hagués, durant mesos, visitat la casa sense topar-se
amb la Maria
(02 ') * S'havia quasi convertit a la religio de la seva promesa quan....
(03') *? Aisâ hauria indubtablement accelerat elprocés
Most of the above examples are ungrammatical. Considering Suñer's
arguments for positing a cliticization rule, it is relevant to allude to the fact
that Catalan existential "haver" does not parallel to Spanish existential
"haber": in Catalan, all the forms of the verb require the locative clitic "hi",
which has survived in modern Catalan.
(09) Hi vaig
(90) Hi ha molta gent
(91 ) Hi havia molta gent
Clearly, then, if we follow Suñer's argument - i.e. the reference to the
(no longer) clitic "y" as evidence for the necessary cliticization of the form
ha in sequences of A?+ past participle -, all the forms of the verb haver in
Catalan must cliticize; in the hi + haver sequences, it is hi which cliticizes
onto the lexical head, haver; in the haver* past participle sequences, it is
the haysr form which cliticizes onto the participle - independently of the
number of syllables it has .
As has been pointed out in 3.3.1.3, an explanation to the fact that
interruption is not allowed in Catalan under any circumstance, but
marginally accepted in Spanish, may be that although incorporation occurs
in all forms of complex verbs both in Spanish and in Catalan, subsequent
cliticization may fail to apply, allowing for a certain degree of non-
cohesiveness. Nonetheless, even in sequences where cliticization may be
argued to have failed to apply, the acceptability of the sentence is not
complete - cf. (72)-(ô3) • Incorporation, thus, accounts for the marginality
for most speakers of Spanish in the Iberian Peninsula of Suner's examples.
Italian
Belletti (1990) considers the possibility of the following word order
possibilities in Italian of Neg particle-Aux-PParticiple-Weg adverb in Italian
- the examples in (93) clearly contrast with Catalan and Spanish -:
(92 )a. Gianni sos napariatopJù
b. Maria sosêuscitamai
c. (Qu&llavoro )Nos J'no finito ancora ( 1 )
(93) a. Gianni sos na pió pariato
b. MarianosêmaJusata
c. iVu$iJavoro)NosJnoascora finito (2) *
Belletti makes use of a more complex structure where the past
participle and the auxiliary are both dominated by independent non-lexical
(functional) nodes25. There is also a NegP which includes a Spec position,
the position where the negative adverbs are generated. The movement of
both , the negative particle and the tensed verb to the highest inflectional
node -AGR - would derive the (93) order:
(94) AGRP
AGR NegP
pm Neg"
A Neg TP
\
non
AuxP
Aux
ttò) V
V (10)
Belletti considers but invalidates the derivation of the order in (92)
by a process of incorporation -as the one postulated in section 3-3-1-3 for
Catalan- of the past participle to the auxiliary, and subsequent movement
of both to the highest AGR node. Instead, Belletti proposes that negative
adverbs may be generated in VP-spec position, and that the participle
moves past the adverb to its AGR position , and the Aux element to its
corresponding AGR node. Adjacency is, thus, obtained without recourse to
Baker"s incorporation. ( Belletti ( 1990);(12))
(95) AGRP
Gianni AGR
AuxP
pres Aux AGRP
avère
AGR'
AGR VP
-to più VP
V
•paria-
The choice of one proposal over the other is done on empirical
grounds. There seem to be two very different predictions made by
the two analyses:
" ... given a sequence "NP Aux Adv PstPrt" , whatever the nature of the adverb involved, the incorporation hypothesis predicts that the order "NP Aux PstPrt Adv " will always be available as well, no matter which position the adverb fills, provided that it is a position lower than the (highest) AGR head. On the other hand, if no process of "Aux+PstPrf incorporation is assumed to be available, the prediction is that the order "NP Aux PstPrt Adv " can only be obtained if the adverb in question fills the VP initial position; if it fills any position higher than VP, the final order of constituents will always be "NP Aux Adv PstPrt '*. (p.30)
These predictions are borne out in Italian, but not in Catalan nor in
Spanish, as was shown in preceding sections, as there is no possibility of
"NP Aux Adv PstPrt" sequence, equivalent to the Italian (96), where a
sentential adverb is used in order to prove that, the incorporation analysis
fails given the ungrammaticality of (97):
(96) a. Gianni na probabilmente telefónate
b. Maria è evidentemente partita (14)
(97)a. *Gianina telefónateprebabilmente
b. * Maria è partita evidentemente (15)
A note on Frenen and English
These two languages contrast sharply with Catalan and Spanish in not
allowing adjacency of Aux and Past Participle in certain constructions, for
instance, V movement and Placement of the negative particle, or
Interruption by the negative adverb neverz/amals - this was already seen
in the presentation of the tests in section 2.1 for English- as (96) -(99) and
(100)-( 101) show:
(9Ô)a. Has Peter seen the film?
b. * Has seen the film Peter?
(99)a. Pierre, a-t.-ll vu le film?
b. *A vu Pierre le film?
(lOO)a Peter has sot. seen the film
b. * Peter not has seen the film
( 101 )a. Pierre n a pas vu le film
b. * Pierre n'a vu pas le film
( 102 )a. /have never seen anything as terrible
b. *fhave seen never anything as terrible
( 103)a. Je n al jamais vu une chose aussi atroce
b. *Je n'ai vu/amals une chose aussi atroce
These examples imply the non-application of incorporation in
equivalent complex structures in French and English - cf. Pollock (1967)
for an explanation of such contrasts -. As will become evident in section
3.4, these differences will also lead to the assumption that different kinds
of auxiliaries - N-aux / T-aux in G&H (I960)* s framework - may have
different properties in different languages. HAVE, for instance - a neutral
auxiliary - cf. section 2.4 - triggers obligatory incorporation in Catalan and
Spanish but not in English, although, as all auxiliaries, it T -marks its VP
complement.
These differences seem to me best captured in terms of the general
loss of '•content" ("possess") lexical status of àaver/üatef in Catalan and
Spanish, as opposed to English and French, -where Jiave/avojr have
retained their "content" lexical status in structures other than complex verb
sequences. This is another argument in favour of the loss of syntactic
autonomy of úaver / Jsaber, and, thus, the triggering of process of
incorporation in these two languages as opposed, again, to French and
English. English and French hav$ /avoir, despite their ability to function as
auxiliaries, do not trigger incorporation,2 6
3.3-2 Incorporation and complex predicates
3-3-2.0 Introduction
This section presents the analysis that Baker assumes for those
sequences analyzed in section 2.3 as complex predicates; i.e. verbal
sequences that do not behave as lexical units, but which are a semantic
unit, although there are reasons to postulate a biclausal status of the
structure in which they occur; i.e. Control or ECM structures. I here present
Baker"s analysis as objectively as possible, leaving possible dubious aspects
for section 3.4. His mechanism of VP to Spec CP is one of these dubious
aspects, and I will attempt to render it unnecessary by making use of
G&H's analysis - as presented in section 2.4 -.
3.3.2.1 A note on complex predicates
Reconsider some of the tests presented in both section 2.1 for
Catalan and Spanish, which seem to indicate that the two verbs in the
sequence - as opposed to those in the sequences considered in 3.3.1 - are
not lexical units:27
Catalan:
1. Interruption by lexical elements (adverbs and parentheticals):
(Test 10)
(104) a. Sempre feia portar lesmaletes alseu marit
\>{l)Feia sempre portar les maletos al sou marit
( 105) a. Comença a protostar gitan te gana
b.(?) Comença sempre a protestar quan te gana
(106) a. Molt sovint vol anar a la platja
b.(?) Vol molt sovint anar a la plat/a
(104) a. Podem sortir on barca si voleu
b. Podem, si voleu, sortiren barca
From Espinal (I960). The * marks the positions where the adverb
òbviament" ( obviously) may appear:
( 10Ô). * On Joan * va haver * däbandonar
(109). * la fortalesa *ha estat * essent * observada
(110). *" En Joan *devia * haver vingut
2. Placement oí the negative particle ( or Tensed negative VP):
(Test 3)
From Picallo (I960):
(111 )£n Jordi pot no haver sortit
(112) Començo a no tenir-ne ganes
(113) En Pep voldria no haver-ho de fer
3. V-0 movement (or Tensed verb preceding subject):
( Test 4)
( 114) a. Solia anar a veure el mar la Maria F
b.(?) Solíala Maria anar a veure elmar?
( 115) a. Pot fer el viatge la Pepa?
b.(?) Pot la Pepa fer el viatge ?
(116) a. Començarà a fer les preguntes l'entrevlstadorP
311
b. (?) Començarà J'entrevistador a fer Jes preguntes?
4. Preposing and postposing:
( Test 11 )
(From Hernanz & Rigau ( iyò4))
(117) a. Solia parlar de Jes seves veJJes amistatsparisenques durant
Jestiu
b. Solia durant Jestiupariar de Jes seves veJJes amistats
parisenques
(From Espinal I960, who does not include ?).
( 11 ô)(?) Crèiem que millorava i empitjorant anava
( 119)(?) Van dir que seria empresonat i empresonat va ser
Spanish
1. Interruption by lexical items (adverbs and parentheticals):
(Test 10)
( 12 0)(?) Hacia siempre JJevar JasmaJetas a su marido
(121)(?) Comienza siempre a protestar cuando tiene Èam&re
(122 )(?) Quiere .cada domingo.Jra JapJaya
( 123)(?) Podemos, si queréis, saiiren barco
2. Placement of the negative particle( or Tensed negative VP):
(Test 3)
(124X?) Carmen deo-eria no Jeer tantas noveJas
( 125)(?) J&te alumno puede no tener que hacer el eissmen,
si su trabajo está bien hecho
\. yO. movement ( or Tensed verb preceding subject):
( Test 4 )
( i 2 ÓJÍ?) ¿Puede tu amigo venir un momento?
f 12 7)(?) ¿ Quieren sus padres conocer a su novia F
4. Preposing and postposing:
( Test 11 )
(From H&RO 964):
( 12ô) a. Bebí comentar este lamentable Incidente hace mucho tiempo
b. Debí hace mucho tiempo cementar este lamentable Incidente
(129) (?) Sabíamos que volverla a fumar v a fumar volvió
The fact that move-alpha may break the sequence of two abstractly
incorporated or reanalyzed - cf. below for an explanation - items is implied
by Baker in several passages of Baker (19ÔÔ). It is important to mention
that this is noted in cases where non-syntactic incorporation takes place -
cf. below for an explanation, section 3.3.2.2 -, as, according to Baker, in the
cases of complex predicates in Romance. In his analysis oí copular passive -
cf. Note (32) - he assumes that BE is generated in INFL and that the V
incorporates abstractly into it, but if BE is a V subcategorizing for another
VP - cf. section 1.2.2 Chomsky 1966b - "we may assume that first the main
V reanalyzes with the auxiliary, and then the auxiliary overtly
incorporates into INFL (p.477 in9).2°" In another passage of the book, which
Baker links to the above statement himself, a parallel process is assumed to
take place in the analysis of causatives in Chimwiimni and Chamorro: "First,
the verb reanalyzes (i.e. is coindexed) with the head of its NP object, thus
freeing the object from the need to get Case. The verb tft$n may move to
INFL C, and ultimately to the matrix verb without taking the object NP
along." (p. 279) What these passages imply is that a sequence of two
reanalyzed items may be broken up by move-alpha. In this case, move-
alpha is an instance of a specific type of abstract incorporation, as will be
explained below.
In the examples above we see that the sequence of two verbs may
be broken up either by the interruption of lexical items or by movement (
preposing or postposing , Vo to C ). Even if we take into account the fact
that for some speakers these configurations do not render a wholly
acceptable result, this set of verbal sequences still clearly constrasts with
the complex verbs analyzed in the previous section in that they form a
quasi - indivisible unit, as opposed to an utterly indivisible unit.
Note that although Baker's theory allows for interruption of verbal
sequences - and , thus, makes correct predictions for the examples above
considered - the dubiousness of the mechanism proposed for complex
predicates calls for reconsideration -cf. section 3.4 -.
3.3-2.2 Romance causatives
Baker extends his analysis of causative structures in Chichewa,
Malayalam, etc. to Italian. He posits that the rule of incorporation as
outlined in section 3.2 is not only available in these languages; Italian
seems to have a similar rule. In other words, among all the possible
variation within causative conductions that Baker studies, Italian - and
other Romance languages - group together with Chichewa2 9 and
Malayalam. Note that following the assumptions in section 3.2 the
configurations in both of the following sections are instances of biclausal
structures - they are considered GF-changing processes .
Baker's analysis of Italian causatives is a reformulation of previous
analyses which posited a rule of reanalysis - cf. Section 2.3.1.4, Manzini
(19Ô3)- for such structures. Reanalysis in the actual framework can be
seen as an instance of incorporation at another level of the grammar; i.e LF.
Italian causatives such as ( 1 ) and (2 ):
( 130) Maria fa iavoraro Giovanni ( 121 )a.
(131 )Maria ha fatîo ripararo ia maccftina a Giovanni ( 121 )b.
are analyzed as instances of this type of incorporation. The crucial
fact is that there are basic syntactic properties that these constructions
share with the relevant constructions in the other two languages. If both
may be explained with one general "mechanism for which , according to
Baker, there is independent evidence, the theory will gain explanatory
adequacy. Baker focuses on the characteristics of the NPs following the
causative sequences: the embedded object of a transitive verb, and the
embedded subject of an intransitive verb behave as the object of the main
verb after causativization, and the embedded subject of a transitive verb
bears an oblique case, namely dative - cf. also sections in 2.3.1 for other
analyses -. The following rule is the descriptive statement of this fact:
(132) Causative Rule 1
GF in embedded clause GF in surface clause
ergative oblique (10)
absolutive direct object (41)
where absolutive= subject of intransitive, object of transitive
ergative= subject of transitive
The relevant tests for Italian are cliticization30, which shows that the
two absolutive NPs may be object clitics of the higher verb (133), and
passivization, which allows again both these NPs as subjects (134)3*
(133)a. Maria Jo faJavoraro
b. Maria Ja fa ripararo a Giovanni ( 12 2 )a. and b
( 134)a. Giovanni è statu fatto Javoraro (moJto)
b. la macdtina fu fatta ripararo a Giovanni ( 12 3) a. and b.
These descriptive observations are predicted by Baker's analysis.
They are, thus, explained if incorporation is accepted as a universal
mechanism. The differences these languages exhibit - " ... one important
difference between the two: from the viewpoint of morphology, the
causative verb and the embedded verb are still two separate words in
Romance." p.201 - seem to be reducible to a parameter; i.e. the level of
application of this rule in the grammar. If incorporation in these
(reanalysis) constructions is assumed to take place from S-structure to LF -
and not from D-structure to S-structure as posited for Chichewa and
Malayalam - the differences follow. This, as mentioned by Baker, is parallel
to the well-known parametric variation in Chinese where wh-movement
aplies at LF.
The differences alluded to above are facts which will gain
importance in what follows. One obvious difference that (135) and (136)
show if compared to (130) and (131) is that causativization in Chichewa
and Malayalam creates a morphologically complex unit, one word:
(135) AnayanJ a-na-wa-m&ny-$îs-a ana kwa buJuzi
baboons SP-PAST-OP-hit-CAUS-ASP children to lizard
" The baboons made the lizard hit the children"
(Chichewa) (94)a.
(136) Amma kuWy&-kkonta annay& mill-ice-u
mother child-ACC with elephant-ACC pinch-CAUS-PAST
"Mother made the child pinch the elephant"
(Malayalam) (9ô)a.
(135) and (136) also show that, in these two languages, there is only
one inflectional ending in the causative verbal units. This contrasts with
Italian where each verbal element carries a specific inflection. Crucially,
though, it is the first verb in the sequence, the causative verb proper that
carries the tense and agreement inflection, the lower verb is in its
infinitival form.
The result of this process - i.e. LF-incorporation - is that the syntactic
properties of causative constructions in both groups of languages - i.e.
Chichewa and Malayalam on the one hand and Italian on the other - are the
same, but they allow for the surface structure differences pointed out
above: "... we must give an account of Romance causatives in which they
have exactly the same syntax as (say) Chichewa causatives, but they differ
with respect to morphology " (p. 202) - . The formal characterization of the
contrast between the two groups of languages is as follows. Note, though,
that the relationshiD between the two verbal heads is and must be a.
essentially the same in all of these constructions: the link that the first
verbal element bears with the second one is translated into a government
relation . Coindexing between nodes is interpreted in these constructions as
the coindexing between a part of a complex word and its trace as in the
following. Note that both a. and b. involve a government relationship and,
thus, further incorporation is legitimated - in b. -:
(1375a. IYP. . . (Xi + Yli . . . !XP t i . . .]]
b.Iyp. . .Yi. . .IxpXi . . .]] (124)
If the lower verb is not in a position where it is directly governed, it
moves in order to reach such a position - cf. section 3-2 -. Crucially, the
Romance constructions that Baker analyzes all require movement; i.e
movement previous to the incorporation process, because they all constitue
cases of biclausal structures.
Recall that in section 3.2 the issue of the movement of the embedded
V to a position governed by the matrix V prior to incorporation was already
mentioned. Two possibilities were mentioned without giving specific
examples. Consider the following structures, which instantiate the two
possibilities:
(13Ô)a. S
NP VP
V CP
1 / \ ai:e VPi IP
ti V NP*
V* NP* NP* r
A I VP
0 ti
(62)a.andb.
These structures show the different movement possibilities of the
verb prior to incorporation of the lower verb to the matrix verb. In
transitive constructions, there are two NPs in each structure which, like all
NPs are subject to the Case Filter. The complex verb resulting after
incorporation, will only be able to assign as many cases as simple verbs in
the language may assign. This is Baker's Case Frame Preservation Principle
(CFPP), which states that " a complex Xo category formed by Incorporation
cannot go beyond the maximum case-assigning properties allowed to a
morphologically simple member of that category in the language" (p.355)
So, if Vs only assign one case, there will be an NP lacking case: in a. the
embedded object, NP*; in b. the embedded subject, NP*. Notice, though, that
this lack of case is not due to the fact that they are ungoverned: both these
NPs are governed by the incorporated V since there are no minimality
barriers, and all the X-max are selected. What plays a crucial role here is
the CFPP. In the b. structure, crucially the causative verb is assumed to
have the lexical property of Complementizer Deletion, otherwise, CP would
be a minimality barrier, containing a distinct head. For both NPs , thus, a
marked type of case assignment is needed.
Consider the Italian example (131) above, repeated here:
( 131 ) Maria ha fatto riparare Ja macctttea a Giovanni (121 )b.
In Italian the insertion of a preposition allows the NP subjects to pass
the Case Filter - cf. also section 2.3.1, Kayne (1975), Burzio (1966) -. It is
the NP subject, not the object which surfaces as an oblique in these
constructions, so their surface structure follows by assuming that they are
derived via VP-to-Spec-CP movement, as in ( 13Ô)b.
Chichewa and Malayalam have this same option for embedded
subjects. Notice that these surface with an oblique case in (135) and (136):
A'wa jbu/usftto lizard; Chichewa),
kuttye-kkonta (child-ACC with; Malayalam, SOV).
The derivation in 03ô)a., VI, is assumed for languages with verbs
that have the ability to assign two accusative cases. One such language is
another Bantu language, Kinyarwanda. In Kinyarwanda, Vs assign
accusative case to two postverbal NPs, and both of these show direct object
properties. In the derivation of (13Ô)a., where the lower V has
incorporated, V* governs both of the NPs, the subject and the object,- there
are no "distinct" heads, no minimality barriers-, and since the V has the
capability of assigning two accusative cases, there is no need to have
recourse to a marked type of case assignment. The following example
illustrates this:
(13*9). Umugabo a -r - uuba}: -JJst -a abaantu Jnsu
man SP-PRES-build-CAUS-ASP people house
"The man is making the people build the house." (67)b
(140)
NP
V
VP
CP
V C IP V*
baildj make tj NP* I
VP
ti V NP*
(66)
Going back to Baker's explanation for Italian, Chichewa and
Malayalam causatives (i.e., 130b) it is assumed that only the embedded Vs
of the last two languages undergo movement in the syntax: they surface as
one morphologically complex form. What Baker posits for Italian is a
process of Abstract Incorporation. Incorporation of the lower V to the
higher V takes place at LF. Before this, the V must move to a position
where it can reanalyze with the higher V : Quoting Baker: " Because of the
presence of the Infi node in the sentential object the verb must undergo
movement internal to the clause in order to get into a position to Reanalyze.
This much happens in the syntax by S-structure. [...] if the lower verb is
transitive, the entire VP must move into sentence initial position, so that
the lower object does not violate the Case Filter. " (p. 203). Thus, it is
assumed that for transitive verbs, the whole VP moves to the Spec of CP (as
in (9)b. - cf. also section 2.3.1.2 for another VP-movement analysis of
causatives, Burzio C1966) -. The object NP is assigned accusative case by the
complex V, and the embedded subject receives case via a preposition
insertion rule.
As implied by the quotation, for intransitive Vs, there is no reason to
assume that the whole VP must move, since the NP-subject in the lower
clause will be governed if the V moves head-to-head. The NP-subject is the
only NP needing case and it can be assigned case directly by the complex
verb.
3.3.2.3 Restructuring constructions
Baker's analysis of restructuring - cf. 2.3.2- constructions in Romance
is parallel to his analysis of causatives . Namely, he proposes that this other
type of complex predicate may also be analyzed as an instance of VI, of
abstract VI. What causatives and restructuring constructions share is that
they are made up of two verbs which function as one, a complex predicate,
and, according to Baker, their D-structure is biclausal; they both have a
lower subject and a lower object, but there are some crucial differences/The
embedded subject in restructuring constructions, as opposed to causatives,
must be phonologically null, and it must be preferential with the matrix
subject - as noted and explained in section 2.3. -. The structure is , thus,
analyzed as an instance of Control. This construction is a point of interest
for Baker in his analysis of GF changing patterns since it also involves a GF
change as the following rule describes:
( í 4 i ) Initial &F Final OF
embedded object object
embedded subject X
matrix subject subject ( 12 5)p.2 04
This rule is assumed to coexist with a specific causative pattern.
Baker analyzes the differences between these two types of GF changing
processes in the same way, but alluding to a specific property that
causative verbs have that restructuring verbs do not have: the lexical
property of Complementizer Deletion. This allows the government of the
embedded subject by the complex head V-0 containing a causative V and
an incorporated V, but blocks government of the embedded subject by the
complex head V-0 containing a restructuring V and an incorporated V.
Recall that if there is C-deletion, CP will not be a minimality barrier since C
wil not be a distinct head. The result, if C-deletion does not take place, is
that the only element allowed in this subject position is PRO, as the theories
of Binding and Control ensure. The' parallel boils down to the difference
between ECM and Control verbs.
The structure Baker assumes for the causative pattern found in
Italian is ( 13ôb) repeated here:
U36)b. S
/ \ NP VP
/ \ V CP
/ \ VPi IP
A V NP*
NP* T
/ \ I VP
I ti
As mentioned, in causative structures of transitive verbs the second
NP is case-marked by a special device which assigns it oblique case ; in
restructuring constructions the second NP does not need case, moreover, it
may not be governed because it is PRO. The fact that the second NP could
not get case followed from the Case Frame PreservatJon PrmdpM which
disallows the assignment of more cases by a complex X-0 than are typically
assigned by a non-complex member of the given X category. 32
A
3.4 Incorporation and T- theory
3.4.0 Introduction
The aim of the next three sections 3.4.1, 3.4.2,and 3.4.3 - is to explore
some of the possible interactions of I(ncorporation)-tlieory (Baker 19ÔÔ)
and T-theory (Guéron & Hoekstra (I960)). Since both, I-theory - the part
of it which is referred to in this thesis, and which constitutes its main
hypothesis - and T-theory are attempts to explain the cohesiveness of
verbs in a sequence ; i.e. of complex predicates and complex verbs, I
believe it worthwhile to attempt a look at the extent of compatibility and
interactions that there are between the twTo.
In section 3-4.1 the basic idea that syntactic incorporation ( as
postulated in 3-3-1- for complex verbs ) - a derivation at S-structure - and
T-theory ( cf. section 2.4) - which crucially involves the FDC, an
interpretation principle of LF - are compatible will be approached.
In section 3-4.2 the two LF mechanisms in each theory - LF-
incorporation and the FDC - are compared. Some dubious aspects of LF -
incorporation and the generality of the FDC seem to malee the former
unnecesssary.
Finally, section 3.4.3 is an attempt to rejoin the idea put forward in
section 3.3.I that V [ VP ] subcategorization requires obligatory
incorporation with the notion of neutral auxiliary (N-aux). The questions
raised are: Do only N-auxiliaries trigger obligatory incorporation of their V
complement ? ; Is it a universal'property of N-auxiliaries or is there
language variation?; May the behaviour of complex predicates be explained
by assuming that the first verb of a complex predicate sequence is an
auxiliary , moreover a T-auxiliary, which does not trigger incorporation?
The special status of epistemic modals is also raised.
In this introduction I would like to briefly retrace some notions
which were raised in the "digest" - section 2.3 -, some of which may be - if
only - "relatable" to some of the notions proposed by Guéron & Hoekstra
(19ÔÔ) ; notions like T-chain, the concept of auxiliary as a T-marker, and
the proposal that the first verbal element in a complex predicate sequence
may be an auxiliary in certain configurations.
Rouveret & Vergnaud (19Ô0) - section 2.3.1.1 - posit the notion of
superscript , borne by both, [- N] elements and [ + N ] elements. In the
former case, one may see how verbs in a sequence were already seen as
bearing "indices" of some sort. They proposed this for the French causative
faire - a much trodden on verbal element - as in their rule I repeated here:
(142) Rule I
FAIRER [ - N ]P NP * 1 [ - N M 3
1 2 3
CONDITIONS : (i) [ - N ] does not branch
(ii) NP is the Theme of I - N 1
The fact that Burzio (1966) - section 2.3.1.2 - proposes VP
subcategorization for certain - obviously - non-derived causative structures
also relates to G&H's proposal of faire as an auxiliary, although - cf. 2.4 and
below - G&H assume IP in the syntax. In the parallel structure analysis of
Zubizarreta - section 2.3.1.3 -, we find that faire is generated as (part of a)
a complex verb in one of its parallel simultaneous structures; another link
with its auxiliary status in G&H. The fact that Manzini (19Ô3t>) - section
2.3.I.4 - proposes the lexical property of renalyser for a causative like
faire, again indicates its - at obvious different stages of the model -
theoretical bond with G&H's proposal.
The proposals for restructuring constructions offer no parallel
contrast possibilities , although Riz2i (1952a) " complex verb" formation -
and all the other proposals which obviously posit mechanisms for linking
M-A + infinitive verbal sequences - may be subsumed by G&H's proposal of
classifying certain of these verbs ( in specific constructions ) as auxiliaries ,
which, therefore, T- mark their VP complements. As will be seen in the
next sections, VP-movement is rendered unneccessary by T-marking, when
the verbs have an auxiliary status; otherwise - if they have a lexical verb
status - they Case-mark their complement and it is interpreted as an
argument, not as ( part of ) a predicate.
Note that in G&H's terms, the notion of complex verb and complex
predicate may be unified at LF by their mechanism of FDC - cf. (143)
below-: verbs forming a complex verb sequence are always part of a
predicate, and when complex predicates are formed , verbs in these
sequences are also part of a complex predicate. Evidently, this is so because
their aim is to contrast between argument/predicate ( nominal/verbal )
verbal complementation, which is not the aim of this thesis. It must be
noted that in order to avoid confusion whenever I refer to the notion of
comptez pr&tieate as in G&H09ÔÔ) - as a result of the FDC -, I will use
italics to distinguish it from the notion of complex predicate used in this
thesis - as explained in Chapter 2 -.
3.4.1. Syntactic Incorporation and T-theory
3.4.1.1 Another look at T- theory 33
The aim of Guéron and Hoekstra to consider the concept of auxiliary
more closely is, to my mind, in the sense of Wass's passage - cf.
introductory quotation to 2.1 - :* Un élément portant le nom d "auxiliaire ne
peut se justifier que s"il permet de révéler des propriétés significatives du
langage ". By giving the notion of auxiliary a firm basis - that of T-marking,
essentially - G&H (I960), to my mind, achieve a simplification of the
grammar in that they allow for VP complementation in cases where
traditional arguments for Control or ECM structures would not allow a VP,
and , thus, explain the exceptional behaviour of complex predicates, and
also, more obviously, the behaviour of complex verbs. This is done by
making intensive use of the level of LF, which is where VP
complementation of otherwise not straightforward VP projections is
caBstru&d Their framework gives a basis - in terms of a construal
principle -, which allows for a specific interpretation of an XP complement
as nominal or verbal - argument or predicate -. The FDC, repeated here:
(14^) Functional Determination of Categories
a. External.
An XP is construed as a nominal projection iff it is
casemarked.
An XP is construed as a verbal projection iff it is T-marked.
b. Internal.
The subject of a nominal projection receives a Case which is
determined internal to XP; the subject of a verbal projection
receives a Case ( if any ) which is determined by an external
governor.
ensures that the complement of the V in a complex verb structure is
construed as part of a complex predicate. Note that I will only refer to the
external determination of XP complements; i.e. I will focus on the notion of
T-marking. The fact that there is no internal subject to the VP of a complex
verb makes it dependent as a part of a comptes predicate. Quoting G&H :
"Arguments, then, are Case-licensed. Their independence is internally
brought out by the fact that if they have an internal subject, this subject is
licensed independently from the syntactic context in which it appears .[...]
In contrast, if the subject of a complement is licensed from toe outside, the
complement is in that sense not entirely autonomous." (p.36) - see section
3.4.2.2 and 3.4.3 for external licensing of subjects in complex predicate
structures -.
How is the XP complement of a complex verb determined as verbal
by the FDC? Recall - cf. section 2.4 - that either Tense or auxiliary verbs
may T-mark; the latter also have the ability to pass on a T-index assigned
by Tense or another auxiliary. It is assumed that an auxiliary always
governs the tense morpheme of its non-finite VP complement. This is a
basic distinctive feature of auxiliary verbs versus lexical verbs, the latter
do not have the ability of T-marking - or passing on a T-index-. In a
complex verb structure, Tense assigns a T-index to the auxiliary , which
then passes it on to VP of the main verb and the main verb absorbs it. This
is, in essence, the formation of a T-chain in a complex verb structure, as
exemplified in (144)a. (022)b. in G&H))- b. for French and English, where
the question of incorporation does not arise; cf. 3-31-4 and 3.4.3.2 -:34
(144) a. Jean { TA' a/Â' [ vpk ¿iÀ' chanté ¡Il
t>.JonnfT*nasj¿{yp£ tj¿sungjf/
Note crucially that, despite the application oí incorporation, which I
am assuming is - in complex verb sequences - an S-structure phenomenon,
the verbal projection will nevertheless be construed as verbal because of
T-marking - rather the passing on of the T-index - of the auxiliary in such a
sequence. Hence, in complex verb sequences verb movement of the lexical
verb - i.e. incorporation- is not needed to identify the XP complement as a
VP, as in the following structures, the Catalan and Spanish traslations of
(3)a. and b. -1 have ommitted showing the application of incorporation, cf.
3.4.1.2 , to illustrate this fact - :
(145) a. En Joan (T* í yp^ha* í ypk cantat fII
b. Juan (T& ÍVPÁ'nak í VPA' cantado III
G&H point out that there are constructions where verb movement is
needed - although in such cases it is movement of an auxiliary and not a
lexical verb - as I am assuming for Catalan and Spanish in the structures
for which I posit incorporation - to identify the complement of a verb as
verbal, and, consequently, to license it. In Portuguese tensed infinitival
examples like (146) the equivalent of the "Aux-to-Comp" - cf. Rizzi (1962c)
and, as already explained in 2.4 - phenomenon ensures the identification of
the SP as a VP at LF. In the following example, the verb lamentar is
assumed to have an optional Case to assign, if it does not assign it - as in
(146), the CP complement must be T-marked in order to be licensed - as it
is not assigned Case -, but since lamentar is a lexical verb, auxiliary terem
must raise in order to T-mark the complement and, thus, identify it as a
verbal projection at LF - and, thus, license it -. ( (9)a.,b. G&HU9ÔÔ)):
(146) a. Fulamente fcp teremjI¡posdiputadostj
[ yp trabalhado pouco JIJ
b. Eu lamente [yp-] teremj fvP-2 os deputados tj
f yp.j traùalhado pouce III
As was explained in section 2.4, the segments in (146)b. are
construed as part of the same VP, following Chomsky (1906b). The
assignment of Case to the subject was also explained in section 2.4 as a
consequence of government by teres? - i.e. an external governor -.
There are several aspects of T-marking and the subsequent creation
of T-chains35 that are closely related to complex verb structure of the type
studied in 3.3.1 - in section 3.4.2. we will see the implications of some of
these aspects for complex predicates -. Recall that T-marking implies the
passing on of a T-index, which percolates to the semantic head of the main
verb VP and becomes part of its reference. Auxiliaries, having no
referential value, cannot integrate a T-index, so they pass it on until there
is a VP headed by a main verb which may integrate it, hence, absorb it.
Another aspect of T-chain formation is that it provides a way for certain
syntactic features to percolate along it. As was illustrated in 2.4, for
instance, theta-role percolation (147), and (148); - (136),(137)in G&H09ÔÔ)
-, which is assumed to be almost identical in simple tenses consisting only
in one verb form, such as ( 147), as in complex tenses, such as (146):
(147) a. John saw Mary
b. John s fagrJ Tlyp saw Mary //
theta-l
( 140) a. John has seen Mary
b. John * fagrJ T[ ypj has [ yp2 seen Mary II
theta-1
The equivalent examples in Catalan and Spanish presumably imply
the same percolation of features by T-chain formation -although the
Catalan simple past example implies a complex verb sequence (Ô) -:
(149) a. En Joan va veuf'? Ja Maria
b. En Joan J [ agrJ T[ VPJ va ÍVP2 veure Ja María ///
thêta-1
( 150) a. Juan * { agr1 Tíyp vio a María I !
thêta-1
Another of the crucial facts about T-chain formation is that they may
be extended by the application of Spec-head agreement (SHAG) or head-
head agreemeent - i.e. when the two V heads share a T-index . This
explains that NP-raising and clitic climbing are only allowed in
configurations where the complement is interpreted as verbal in LF,
because it is T-marked. As explained in 2.4, NP-raising and clitic climbing
are possible when antecedent government holds - as in the passive account
in Barriers-cf. 1.2.2 - : VP is nota barrier because of Spec-head agreement
(SHAG); I shares index with the governed ec by virtue of raising or
"climbing" - cf. 2.4 -, and the ec is antecedent governed, as in the following
complex verb structure with clitic climbing- (76)a. in G&H09ÔÔ) -:3&
(15D Je J j ailypej vu ejl
It is obvious, thus, that in complex verb structures, clitic climbing
should always be possible, as is indeed borne out by the facts, and
independently predicted by G&H's framework. - as quoted in section 3.4.2.2,
(23) -3?
WDLj-Hef Cejjvist oj/
(153) Hoj vaig [for OÍ I
(154) Lsj ho[ visto oj/
3.4.1.2 Complex verb incorporation and the FDC
There is a basic compatibility of syntactic incorporation and T-theory,
with its basic construal principle at LF, the FDC. As pointed out in the
preceding section, the fact that complex verbs are made up of a sequence of
verbs, the first of which is an auxiliary ensures previous T-marking, and ,
thus, only one possibility of construal: the XP complements of the first verb
in the sequence in complex verbs are always interpreted as part of a
comptes predicate
Syntactic V-movement in incorporation structures - as explained in
3.2 - requires antecedent government of the V trace. This is achieved, as
explained, by S-structure, so the LF construal principle, the FDC, should not
affect this relationship. The following structure shows that no principles are
violated, if we assume incorporation to have applied plus the FDC: VP-2 is
T-marked, and, thus, licensed as a VP, and the trace of V-2 satisfies the ECP
by antecedent government :
(155) VP-l
I V-l
/ \
V*i/j VP-2j
V-lj V-2i V-2
V-2
¿i
(where j - as in section 3.4.1.1 - is the T-marking index )
As explained in 3.2, Theta-indexing between the X and the YP is
what devoids the Xmax dominating the trace of the incorporated item of its
barrierhood status, and, thus, allows antecedent government of the X-0
trace by the incorporated V*. In complex verb incorporation an equivalent
mechanism - i.e. either Chomsky(19ô6b) L-marking, or Balier (I960) theta -
indexing - must be assumed. Nevertheless, in G&H' s framework there is
another possibility, which I will assume: they assume that T-marking is
equivalent to L- marking. There is, thus, no need to recur to a special type
of theta-indexing in complex ver.bs in order to devoid the VP of its
barrierhood status if it is independently T -marked; antecedent government
of the trace holds. Note, nevertheless, that T-marking is alluded to in an LF
mechanism, the FDC. G&H, though, imply, to my mind, - and even more by
making it equivalent to L-marking - that it holds at S-structure - as does
Case marking, which is also alluded to in the FDC.
It is crucial to distinguish between these sequences of verbs and the
behaviour of complex predicate verbal sequences in that precisely the fact
that syntactic V-movement; i.e. incorporation - a process which I do not
assume for complex predicates, as will be explained in section 3.4.2 - , has
applied accounts for their indivisibility - cf. 2.1 and 3.3.1.1 -. Consider some
more examples that show their indivisibility:
a) Interruption by lexical elements is not allowed:
( 15b)a. *No m Ms mai dit. que tenies tres fills
b. * Va, sense voier, tirar tots eispiatsper terra
(157) * No me nas nunca dicho que tenias tres nips
b) Preposing results in ungrammatical structures:
( 15ö)a. * Ens pensàvem que es casarien però casat no sñan
b. * Crèiem que ia Maria £osa iguaria pero baiiarnc va
(159) *¿ Yeliibro ? - ¡Comido no me io ne/
Another important aspect with regards to the relationship between
S-structure and LF in complex verb sequences is that they may be assumed
to be identical 3Ô: if no functional nodes intervene, the FDC has no effect in
eonstriung a CP/IP as a VP.
3-4.2 LF-incorporation and /or tîie FDC
3.4.2.1 Another look, at LF-incorporation
Complex predicates are quasi-indivisible units - ci. 3.3.2.1 -, tîiey are
thus, structures for which syntactic incorporation cannot be posited. Baker
states that abstract incorporation - or Reanalysis - has the same properties
as movement because Reanalysis constructions form a natural class with
incorporation structures in other languages " whose properties follow from
the theory of movement ", he concludes that " it has the same properties as
movement simply because it too is movement, albeit, movement which one
cannot see." (p.203). If a simpler analysis, which is independently
motivated, explains this unseen movement I believe it must be chosen on
simplicity grounds.
It is also stated in Baker (19ÔÔ) that because of the presence of I in
the embedded sentential complement, the V must undergo movement
internal to the clauses - see section 3.3.2 -, in order to get into a position to
reanalyze. If no such structure is needed, movement will be made
unnecessary - cf. 3-4.2.2. and 3.4.3 -, provided that a mechanism like chain
formation, for instance, is independently needed, and, thus implies no
addition to the theory .
There are some dubious aspects of abstract incorpora tJcn which may
be considered to render it unnecessary as long as the notion of T-marking
of G&H (19ÔÔ) is sufficient to explain the behaviour of complex predicates
and makes the correct predictions. Firstly, the VP-to-Spec CP mechanism is
posited in order to obtain a government relation between higher verb and
lower verb in biclausal structures, without there being a clear trigger for
this movement. It must be noted - cf. also 3.3.2 - that the V-projection
movement is interpreted as VP movement because in causative transitive
structures it is the subject which surfaces as oblique - i.e. preceded by the
preposition t? -. Furthermore, the usual targets for movement into Spec-CP
position are not VPs; i.e. it is usually wh-phrases that move to Spec-CP.
Obviously, this does not imply that a VP may not be subject to
move-alpha, there are obvious cases of VP preposing39, but in these cases,
the landing site for the VP is plausibly not Spec-CP but rather adjunction
positions created as in instances of extraposed constituents, as in: Nobody
would ride with Frod who know him -from Radford (19ÔÔ) P-565; an
instance of S"-adjunction -. (160) may also be analyzed as adjunction, in this
case, of a VP;
(160) Eido with Fr od. I wouldn't if I were you/
Moreover, an LF mechanism such as abstract incorporation is
questionable if reference must be made to an S-structure condition such as
Case. Baker refers to the Case Frame Preservation Principle (CPP) - cf.
3.3.2.2 -, which makes the proposal dubious crucially for Romance
languages, where incorporation is assumed abstract. VP-to-Spec-CP may
not be questionable for languages where there is syntactic VI -
incorporation. Note that Baker himself points out this problem in a footnote:
(161) " There is one important problem with this suggestion,
however. Given the standard view of GB, all "overt" movements
that occur between D-structure and S-structure are assumed to
strictly precede all "covert" movements, which happen
between S-structure and LF. Yet, I will have cause to claim at
various points in what follows that covert Incorporation
crucially precedes overt Incorporation ( or seems to ) in certain
cases, giving rise to ordering paradoxes. This is even seen
directly below where VI "precedes"Case assignment which I
have assumedhappens at ¿structure. These facts may imply
that Reanalysis, although abstract incorporation' in some
sense, is not LF Incorporation after all. On the other hand , the
true relationship between LF and the other levels of syntactic
description is a controversial topic and may need to be revised.
[...] "( p. 462 in. 37) (italics mine)
Further related to Case considerations for Romance causatives , a
marked type of case-assignment - the a-insertion rule- is needed to Case-
mark the subject of transitive verbs. The NP direct object is assigned
accusative case by the verb of which it is a complement because the whole
VP moves. As Baker himself points out, there is no need to postulate a VP-
to-Spec-CP derivation for intransitive verbs, where V-movement is
sufficient for the subject to be assigned Case. The prediction is that the a-
insertion rule should not apply in intransitive causative structures. This,
however is not borne out especially in Spanish and also Catalan, as the
following examples of intransitive causative sequences illustrate:
(162) Hizo reir *(a) su hija
(163) Va fer riure (a) la seva filla
This marked type of case-assignment is, thus, independently needed,
whichever mechanism is posited for causatives in Romance languages.
In restructuring constructions in Romance - rule 3 in 3.3.2.3,
repeated here -:
(164) Rule 3 (p. 204 (125) in Baker (19ÔÔ)
Initial GF Final GF
embedded object object
embedded subject 0
matrix subject subject
there seems to be no real GF-changing process; a function
"disappears" but there is no change. This fact, the already noted
dubiousness of VP-to-Spec-CP movement, plus the fact that there is no
phonologically realized subject in "restructuring" (Control) structures, seem
to give even less motivation to posit a VP-to-Spec-CP derivation for these
structures.
34.2.2.CanT-markingreplace LF-incorporation?
" The hypothesis that causative, perception, and, often, modal
verbs function as auxiliaries simplifies the grammar. It obviates
the need for a number of powerful syntactic devices which have
been proposed in the literature essentially in order to eliminate
the CP node of the complement of such verbs." G&H( 19ÔÔ)(p.55)
As already briefly noted in section 2.5, most mechanisms
summarized in Chapter 2 are not wholly valid in the present framework.
G&H (I960) argue against some of these on the basis of : a) their not being
independently motivated; b) their violating constraints in the present
framework; and c) implying an addition of undesirable powerful
mechanisms to the grammar. In the previous section I have attempted to
extend the criticism to Baker"s LF-incorporation mechanism. In this section
I will try to show that T-theory, T-marking and the FDC, make correct
predictions, and, thus, there is no need to allude to abstract incorporation.
Note also that for G&H (i960) - with obviously different aims than Baker
(I960) - the causative and "restructuring" constructions are not regarded as
GF-changing processes as in Baker (I960). The GF-changing process is here
"reinterpreted" as a construal of a complex predicate because of the
special nature of the first verb in the sequence. Quoting G&H: " ... IP is
construed as a nominal projection nondistinct from NP when assigned case
by Infi, V, or P, but as a verbal projection non-distinct from VP when
governed by an auxiliary verb." (p.37). This allows for a propositional
complement- - as in the case of complex predicate structures - to be
construed as either an argument or as part of a comptez predicate at LF
provided that the first verb in a complex predicate sequence may be
granted an auxiliary status in certain configurations.
This paves the way for a different explanation within the present
framework of the special behaviour of verbs that diverge from "typical"
main verbs - as already noted by Rizzi (1982a) and in all the works
summarized in section 2.3 -; i.e. their semantic and syntactic cohesion, plus
the degree of divisibility that they allow - cf. 3-3-2.1 -, which distinguishes
them crucially from complex verb sequences. What they share with
complex verb sequences, nevertheless, is independently predicted by the
FDC, in G&H" s words:
(165) "The rules proposed by previous authors in order to
account for clitic climbing or NP raising in complex structures
derive output configurations which are syntactically equivalent
to base structures headed by auxiliary verbs. These structures
independently allow clitic climbing and NP raising. Moreover,
reanalysis and parallel structure rules do not apply to
auxiliary structures, which contain no embedded CP to begin
with. If, as we propose, "restructuring" causative and modal
verbs are analyzed as auxiliary verbs, the syntactic behaviour
which characterizes auxmarv -headed structures automatical^ * »
applies to structures headed by causative and modal verbs."
(p-50)
To briefly summarize the aspects of complex predicates which will
concern us we may consider a few examples to recall their "double nature"
- cf 2.1.2 and 3-3-2.1 - in order to focus on the relevant aspects for the two
theories that are being made use of in this thesis; namely Baker(19Ôô) and
G8tHil9ôô). On the one hand, complex predicates are syntactically (and
semantically) united: a)(l66) and (167) show their non-interference with
the theta-grid of the main verb:
(166) a. EiMiquoi fa pot caminar
b. Una casa teta do fusta pot cromar-so fàciimont
(167) a. Aianyapuodo cantar aiguna canción
b. Una casa do madorapuodo quomarso faciimonto
b) and clitic climbing is possible over complex predicate structures:
( 16ô) a. la pot utiiitzar
b. Sompro Ja fa proparar ai sou marit (iapaoiia)
( 169)a. Lo quioro vor
b. Siempre se la hace preparar a su marido
On the other hand, the possibility of separating the two components
of a verbal sequence of a comptez predicate, points towards the non-
application of incorporation, as i will assume below - and already noted in
3.4.I.2 -: a) Interruption by lexical elements is marginal, but acceptable:
(170) a.? Fa sempre preparar Ja paella ai seu marit
b.? Vol sovint veure Ja matensapelJícula dues vegades seguides
c. ? Pots, sempre que vulguis, fer servir el meu eotse
( 171 ) a. Le hace siempre planchar la ropa a su marido
b.(?) Quiere a menudo ver la misma película dos veces seguidas
c.(?) Puedes, siempre que qiueras, utilizar mi eoLYe
b) Preposing is also (marginally) accepted:
(172) a. ? Sabiemque voldria anar-hi i anar-hi va voler
b.? freiem que el faria cantar però cantar no el va fer
( 173) a.? Pensábamos que podría ir, pero ir no pudo
b.? Sabíamos que lo haría cantar y cantarlo hizo
Having briefly reviewed their behaviour, an account in terms of
G&H's proposal is due. As already noted, it will be in the following line: *' If
the infinitival is construed as a verbal projection, it is T-marked by the
morpheme of an auxiliary verb , as in causative and modal structures in
many languages ..."(p.73). Note that reference to the auxiliary type to which
the first verbal element belongs is not alluded to in this section - cf. section
3.4.3 for reference to it -; here I will only refer to the possibilities of T-
marking of the first verbal element in certain complex predicate structures.
The account of complex predicate constructions by the FDC is sufficient to
explain the behaviour of complex predicates if everytime a complex
predicate is formed, VP construal is ensured . This would account for their
syntactic and semantic cohesion, - and, thus, possibility of phenomena such
as clitic climbing and NP raising- and, the fact that the FDC is an LF
mechanism would account for their S-structure divisibility possibilities. If
their account is correct, G&H (19ÔÔ) provide the mechanisms for an
infinitival S to be construed as a verbal projection, without making use of
undesirable nor unallowed mechanisms in the present framework of
generative grammar, and they account for both complex verb and predicate
structures.
Hence, even if there may be reasons to posit a biclausal structure -
cf. 2.2.1; basically the arguments for positing a Control or ECM structure -,
in the T-theory framework there is no need to appeal to movement of a
verbal projection - cf. Baker (I960) mainly, but also Rouveret & Vergnaud
(I960), Burzio (1956), as summarized in section 2.3 -, the syntactic and
semantic unity of complex predicates is maintained, because of the FDC,
granting V-1 an auxiliary status.
The above assumptions imply that there is a parallel explanation for
complex verbs and certain complex predicate structures, as the following
examples - G&HU9ÔÔ) (76)a.-c. - illustrate:
( 174) a. Je Jj aJ fej vu ej 1
b. Je Jej ferai! ypJire ej/
c. Loj voglio / yp/eggere ej /
G&H grant a biclausal status to complex predicate constructions in the
syntax, which allows for general principles of the framework, to be
maintained, for instance, that the external theta-role is assigned to the NP
in subject position via INFL - p. 65 (67) -:
Í175) The external theta-roie is assigned to INFL by VP
The Visibility condition wiii aiiow ÁGR in Infi to retain the theta-role
assigned to it - and not transmit it to the subject position as long as it is
assigned Case - cf. below for constructions of modals and causatives where
this is assumed to take place -.
T-marking, as noted, implies the formation of a T-chain, and there
are several aspects of this phenomenon which also involve the
constructions of complex predicates. Most of these coincide with those
noted for complex verbs; i.e. the fact that a path is created for the
percolation of features, and that extension of chains is possible, allowing for
clitic climbing and NP-raising. Crucially, though, the fact that each CP
introduces a new T-index is only relevant for complex predicates, where an
extended chain is only possible if T-marking takes place and CP is
construed as a verbal projection.
Note the following examples of both, a causative and a modal
complex predicate structure, which G&H(19ôÔ) - (95X96) - analyze as VP
construal ; as noted in section 3.4.1.1, the fact that there is participle
agreement implies that an NP or clitic goes through SC subject position
interpreted as VP at LF; i.e. governed by an auxiliary : fare, and volere in
this case.
(176) a. Man'sfu faitsinvitare
b. Mariai {ypi fu ÍVP2 &i fstta fypj invitare ei / / /
(177) a. Marls e voluta tornare a casa b. Mariai ívpj è í VP2 <?/ voluta f ypj tomare ei a casa / / /
Causatives as auxiliaries
G&H posit that French and Italian - the two languages on which they
focus - causatives are auxiliaries, and I will assume the same holds for
Catalan and Spanish. They also assume that English maÉe/let have an
auxiliary status, as will be illustrated below.
As noted, a consequence of the formation of a T-chain is that all the
Vs which intervene between an extracted element and its original position
must be auxiliaries - except the last one -, otherwise, there would be no
passing on of the T-index, and antecedent government would not hold. The
following examples show that this holds for French - G&H (173)a. - Catalan,
and Spanish for clitic climbing, but (Iô2)-(lô4) show that it only holds for
Italian for NP-raising - a fact unexplained in the framework-:
( 176) a. ft lalfalt voir
fc>- ft Í VP1 ii 'si / YP2 fait { ypj voir ei III
(179)a. La vaig fer llegir
b. ÍVP1 Lsi vaig í VP2 fer ( ypj itegir ei III
(100)a. L'lie fet llegir
b. / ypi L l'he [ VP2 fet [ VPJ Hegir ei III
( 1 ô 1 ) a. La ha hecho leer
b. / VPI Lai A? / VP2 he±o [ ypj leer ei III
(1Ô2) a. 1 llbrl furono fattl leggere
b. llibriiypi fnroj2ûfvP2 ejfattifypj ej leggereei///
( 103) *&s llibres van serfetsllegir
( 104) *Loslibrosfueronneones leer ̂ 0,4. 1
The auxiliary status oi the causative faire is claimed to be
independently supported by G&H by several facts. One of these is the fact
that it allows VP anaphora. They cite Zribri-Hertz (1966) in claiming that
auxiliaries are those elements which may carry the tense of a null VP. Note
that this is in contradiction with the examples of section 2.1 and 2.1.2
which show that null VP is allowed by main verbs - (ô5)p. 174 G&H ( 19ÔÔ):
( 105) feao $ Í acnetédupain fee matin et
Pierre lejfera ej ce soir
Catalan and Spanish causatives have the same possibility:
( 156) En Joan Carlesna marxat aquêstmatfJ
la lsabel no farà aquesta tarda
( 167) Juan Alberto cocinó 'morallasayerporla noene
y José Marià lo Mará mañana
These facts, apart from possibly being independent evidence for the
status of faire/fer/nacer as auxiliaries - if we assume G&H's arguments -
are evidence showing that faire is a "support" verb in the three languages
mentioned. Note that for the hypothesis given in this thesis, the facts do not
disclaim it: the only reason why the auxiliary verbs in Catalan and Spanish
do not allow VP anaphora may be explained by the fact that they undergo
obligatory incorporation, as opposed to a verb like the causative, which is
not part of a complex verb, but rather a complex predicate.
Another fact that G&H take into account as evidence for iet being an
auxiliary is that it takes a bare infinitival, supposedly a VP:
(lôô)a. They M /ypNoemaieave ]
b. * They iet Noema to leave
As opposed to French fairs Catalan fer, and Spanish hacer,
make/iet/iaisser are structural case assigners, so they can assign structural
case to the subject of their complement - even if it is contrued as a VP SC -
( 169) We made the students sit in rows
(190) Mous avons iaisséies étudiants sassoiri Iw derrière i'autre
(191) *Nous avons fait ies étudiants sassoirilw derrière i autre
(192) *Hem fet eis estudiants seure i'un darrera i'aitre
(193) * ¿femes hecho ios estudiantes sentarse en fila
Catalan and Spanish also have a causative predicate which allows
structural Case assignment of the subject of its subcategorized clause:
deisar/dejar: Deijca ai teu fifi venir amt> nosaitres //Deja a W hijo vivir
comoquiera.
In 3.4.3.2 this fact will be expanded and explained; the fact that the
theta-roles of the complement of the causative are all assigned within the
complement is a consequence of the type of auxiliary it is, which has a
specific type of complement.
Modals as auxiliaries
Note that there are constructions in which modals function as lexical
verbs and constructions in which they are auxiliaries. As G&H point out,
they have many of the properties of auxiliaries that HAVE and BE have
over languages. In English, they occupy the INFL position and raise to
COMP, like other auxiliaries. No question has been raised as to the auxiliary
status of modals in English. Putting it differently, the fact is that we may
not find equivalent "restructuring" constructions in English as there is no
independent motivation for a "double nature", as there is in the Romance
languages considered here.
G&H allude to Pollock (19Ô7) - cf. Chapter 4 - in claiming that modals
in French are auxiliaries as they raise to INFL, like other auxiliary elements
in the language, and unlike lexical verbs. They reinterpret this in their
framework by granting them the ability of T-marking, thus, of being
auxiliary verbs in certain constructions. - ( 194)a. and b. are ( 19 l)d and f. in
G&H) and the contrast */? is an argument for giving the two verbs, pouvoir
and croire a different status -:
(194) a. ! Il pensait [ jp PRO no pouvoirj [yppastj dormir ici II
b * II pensait [jp PEO no croiroi
/ yppas ti aus histoires ¿to fan tomos / /
Again, the fact that modals may introduce a null VP is given as
evidence for their auxiliary status -(195) is adapted from G&H's ( 192) -:
( 195) Pierre voudrait aixorder le piano
a. mais il ne peut/Veut/doitpas ¡ypie faire ]
b. maisilnepeutsveut/doitpas
c. * mais il ne croit/dédde pas
If the assumption that auxiliaries may be tense carriers in Romance
languages, and, thus, may introduce a null VP; the facts follow since, again,
in the case of modals, incorporation is not assumed to have taken place - as
they are complex predicates, and not complex verbs in the following
examples - :
(196) Es Fere volaría afinar elplano
a. però no pot/vol/VP fer-ho J
b. però nopotMoJ
c. *perànocreu/decideLY
( 197) Podro querría afinar eJpiano
a. pero no quiere/puede í VP hacerlo /
b. pero no quiere/puede
c. * pero no cree /decide
G&H assume that clitic climbing is possible in Italian in modal
structures because their complements may be construed as VPs at LF; if the
clitic remains in its original position, there is no VP construal. The way I
interpret this is that if the modal functions as an auxiliary, - thus, as a T-
marker- there is clitic climbing, and VP construal - by the FDC - takes place.
On the other hand, if the modal 'does not T-mark its complement, but
instead assigns Case to it - i.e. functioning as a lexical verb - then there is
no VP construal, but rather its complement is interpreted as an argument,
instead of a predicate.
Root modals assign an external theta-roie, accusative case F, and
optionally an internal theta-role. On the basis of which element retains
Case F, the basic distinction noted in the "restructuring" literature is made
in G&H's framework. Basically, two options are possible for Case F: a) that
IP retains it, and, thus, that IP is construed as an argument by the FDC; b)
that Case F is absorbed by I and, by Visibility, Agr may retain the external
theta-roie assigned via IHFL. In this case, Ágr is a pronominal. V raises to Ï,
and IP is construed - by the FDC - as a verbal complement. Another
assumption they also make is that a modal must control an ec in their
embedded SC complement. The a) option is illustrated by (195), and b) by
(199)-(193H194)-:
(19Ô) Marioj vuotefjp PWj [j'fyp loggore il libro ] 11
theta-1 theta-1
(199) Marioj vuote {jp leggoroj [ yp tj il libro 11
theta-1 theta- \\
As illustrated and explained in 2.4, the ECP blocks clitic climbing
when the modal does not function as an auxiliary, (196) ; i.e. there are two
different T-chains, so antecedent government does not hold. Whenever
there is clitic climbing, the modal imitions as a T-marker, and by T-chain
formation, the embedded verb antecede-governs the ec. Therefore, the
account is parallel for complex predicate and complex verb structures.
Note, though, that even if there is the possibility of a modal
functioning as a T-marker - an auxiliary -, it is still necessary to have an IP
in the syntax, although it is construed as a VP in LF; an INFL node accounts
for Control in modal structures, where agr is assumed to function as a
syntactic argument.
The lack oí clitic climbing in modal structures in French - cf. also 2.4
- is attributed to the pro-drop parameter. Basically, where Italian has an
inflectional argument position in INFL - finite clauses and modal
structures-, French needs a Spec-IP argument position, which prevents the
formation of a T-chain. The position is syntactically active because it is
subject to a syntactic rule like Control. Hence, the difference with causative
structures, where French allows clitic climbing; the pronominal in AGR is
not syntactically active, it is not an A-position.
Epistemic modals are analyzed in a parallel fashion - but cf. 3.4.3.2
for a different proposal for epistemic modals -; they also have both options
although they do not assign an external theta-role. The AGR in INFL may
absorb Case F of the verb, but in this case, it functions as an expletive
pronoun, not PRO. IP is construed as VP in LF. In French, the modal, again,
does not have the ability of T-marking; it assigns Case to the IP , which is
construed as an argument in LF.
In Catalan and Spanish, modals may also be claimed to have the
possibility of functioning as auxiliaries; i.e. to have the ability of T-marking,
on the basis of the fact that they allow clitic climbing, which by T-chain
formation, ensures antecedent government:
(200) a. La MariaJosep la j vol* [ yp llegir *: ejf
b. En Josep Maria noi deuJ [yp agrentendreJejf
(2 01} a. Juan Carlos Ay quiere ¿Y ypleer¿ ei /
b. Carmen loi debe1 [yp entender1 eil
To sum up, a point in common that LF-incorporation'and T-theory
have as regards complex predicates is that they both posit a biclausal
structure , and, thus, an identical D-structure for all types of complex
predicates. An important difference - which I have used to attempt an
invalidation of LF incorporation in favour of T-theory- is that I-theory
with respect to complex predicates alludes to a specific sort of movement
versus awstnmJ posited in T-theory, which is , to my mind more in line
with the present framework.
3-4.3 Neutral auxiliaries / T-auxiliaries and incorporation
3.4.3.1 Properties of N(eutral)- auxiliaries and T-auxiliaries
All auxiliaries assign a T-index; i.e. T-mark/- or pass on a T-index -
to a non-finite VP complement; nevertheless, there are systematic
properties across and within languages of different auxiliary items that
suggest a classification into basically two different types of auxiliaries:
N(eutral)-auxiliaries and T-auxiliaries. In this section I will review some of
these properties - cf. also 2.4 -. HAVE will be taken as the paradigmatic
example of a Neutral-auxiliary; BE as the corresponding T-auxiliary. The
differences will obviously become important when I reconsider complex
verb and complex predicate structures in section 3.4.3.2 .
N-auxiliaries combine with the tense morpheme of its VP
complement to form a complex tense morpheme, which defines the tense of
the S 42 T-auxiliaries govern a VP with an independent tense morpheme.
G6:H refer to this as the assignment of a T-role. T-auxiliaries, thus, asign T-
roles, whereas N-auxiliaries do not assign T-roles. G&H make a difference
between theta-roles, assigned by lexical verbs, and T-roles assigned by T-
auxiliaries: T-roles give an independent tense value to their complement;
theta-roles give a referential value to their complement. Another difference
between the two auxiliary types is that T-auxiliaries have the possibility of
assigning structural case, whereas N-auxiliaries do not, as the following
example shows-.p.4ô (33)b. - In the following example HAVE is an instance
of a causative construction:
(202) John had fnJsnouse burned I
Note that the same lexical item, in this case HAVE, may be a T-
auxiliary or a N-auxiliary in different contexts. - as already noted in 2.4 -.
The other crucial difference between N-auxiliaries and T-auxiliaries is that
the former allow percolation of the theta-role assigned by the VP
complement main verb - assigned to NP subject by subj-I agreement ;
indirectly by I. A T-auxiliary does not allow percolation of theta-roles
assigned in its complement. This is related to the notion of Complete
Thematic Constituent 43, repeated here - (30) p.4ô -:
(203) An XP which is assigned a T-role is a Complete Thematic
Constituent (CTC): all theta-roles associated with X, the head
of XP, are assigned internal to XP.
The following examples form G&H ( p. 49(35X(36) and(37) ) illustrate
the difference:
(204) John[j-1[vphas[ VPlaughed///
theta-1
(205) fohnj was f $j seen e//
theta-1 theta-2
(206) /eanjest [ ej venu ej/
theta-1
In (204) the theta-role percolates, as indicated by the arrows. In
(205) and (206), all theta-roles are assigned within the complement of the
T-aux, BE.
As a reminder of G&H"s auxiliary restrictions, which will also become
relevant in the next section, I will repeat here the general restrictions on
auxiliaries (p.50, (40):
(207) a. A T-chain may not contain two auxiliaries of the same class,
b. Neutral auxiliaries precede T-auxiliaries.
In the next section I will propose - with the exception of epistemic
modal verbs - that those lexical verbs which may also function as
auxiliaries, those in complex predicate structures, may only function as T'
auxiliaries not as N-auxiliaries.
3.4.3.2 N- auxiliaries and T-auxiliaries in complex verb and complex
predicate structures
The hypothesis that I will propose in this section is that the complex
verb sequences considered in this thesis are made up of neutral auxiliaries
and that neutral auxiliaries in Catalan and Spanish trigger obligatory
incorporation. They are made up of N-aux precisely because these allow
percolation of theta-role from the lower verb to the matrix subject position.
As noted, this is one of the basic properties of N-auxiliaries and a
fundamental property of complex verbs as opposed to complex predicates;
i.e. only complex verbs have one theta-grid - as explained in Chapter 2 ; cf.
"selectional restrictions" test.
G&H point to the fact that Eater but not avoir is always a neutral
auxiliary-(10Ô) p. 70-:
(206) HAVE is solely a neutral auxiliary in Spanish but either a
neutral or a T-auxiliary in French
In French they posit that it is a neutral auxiliary in its temporal
aspect, but a T-auxiliary in - (109) in G&H -:
(209) a. Ele a deuxfrères
b. We a son fils malade
They link this with the fact that in Spanish HAVE may never have a
lexical meaning, as the corresponding translations of the French sentences
show -G&H (110)-:
(210) a. *Ha dosHermanos
b. *Ha su È1 jo enfermo
Note that the consideration that one lexical item may function as one
type or another type of auxiliary in one context or another, plus their
relating it with the fact that it may or may not have a lexical meaning links
up with the main hypothesis in this thesis that N-auxiliaries trigger
obligatory incorporation in Catalan and Spanish but not in English nor
French; in Catalan and Spanish - as noted above - they may never function
as lexical verbs; in English and French, they may. I repeat here an
illustration of a complex verb sequence where the theta-role is assigned to
the NP in subject position via INFL, by the percolation through the N-
auxiliary, HAVE, and the corresponding Catalan and Spanish translations:
(211 ) a. John Ij'Iívp has { yplaughed HI
~^L_-theta-l b. En Joan / / • J [ypha { yp rigut 11
I——theta-1
c./uan fj'If ypha í yp reído //
i theta-1
In preceding sections, I have been giving examples oí another
complex verb sequence in Catalan , which is equivalent to the simple past,
but is made up of a sequence of verbs: anar * infinitive. By the examples
given so far, it must be concluded that this auxiliary also triggers obligatory
incorporation in this sequence, so it should be an N-auxiliary if my
hypothesis is correct. Nevertheless, I have alluded to the lack of lexical
meaning in both Catalan and Spanish of the N- auxiliary HAVE, precisely as
a factor that might contribute to its triggering incorporation - as opposed to
English and French HAVE -. The fact that anar has not lost its lexical
meaning, and yet, it triggers obligatory incorporation may be related to the
different development in Catalan as opposed to other Romance languages of
the verb anar+ infinitive. In Catalan it has developed into a simple past
equivalent - the one form simple past is practically out of use -, whereas in
other Romance languages it has developed into a periphrastic future
aspectual sequence.^
Note that one of the basic properties of an N-auxiliary, apart from
allowing the percolation of the external theta-role of the V heading the
main VP, is that they " combine with the tense morpheme of their
complement to form a complex tense morpheme defining the tense of S ", as
opposed to T-auxtlianes which " govern a VP with an independent tense
morpheme" (p.4ô). An illustration of the fact that this "periphrastic" simple
past allows percolation of the external theta role is that there are no reason
to claim a biclausal structure, as in the example; it combines with the tense
morpheme of its complement :
(2 í 2) En Joan ¡i • /•* / VP va ¿ í VP riure & /
theta-1
The hypothesis in this section as regards complex predicates - when
they arise - is that they are made up of T-auxiliaries and they do not
trigger obligatory incorporation - with the exception of epistemic modal
verbs -. When the first verb in a sequence of a complex predicate functions
as a main verb, its complement is interpreted as an argument, when it is an
auxiliary, its complement is interpreted as part of a aimpJ&Ypr#tâc3te.45
The fact that there are two theta-grids in some sequences of
complex predicates, and these are still construed as VPs at LF is not a
problem for G&H's framework by their introduction of their concept of T-
auxiliary and of CTC - cf. (219) below - plus the possibility of assigning a
theta-role to AGR in INFL - as explained in 3.4.2.2 - Basically, when the
Case assigned by the verb is absorbed by the I, AGR functions as a
pronominal, retaining the external theta-role.
Note that in complex predicate verbal sequences incoporation should
not be posited because their degree of separatibility shows that they are
not an amalgamated unit. Nevertheless, the VP dominating a main verb
may still be devoided of barrierhood by T-marking, so incorporation is
predicted possible. Nevertheless, it is the fact that the complement of a T-
aux in causative and restructuring constructions is construed as a VP at LF
which makes syntactic incorporation unmotivated, thus allowing preposing
and clitic climbing :
(213)a. (VP1 V-1 Icp IIP (VP2 V-2 NP ] ] ] ] S-structure
b.lypiV-llvP2 V-2 NPH LF
At S-structure, V-1 may T-mark VP2, so V-2 can move and NP can
also "climb" out. Interruption by lexical elements is also predicted possible
by the assumption that incorporation does not apply - cf. HOC, section 3-2 ;
there is no "opaque" constituent -j adjunction sites for adverbs are
available.
I will thus assume that in Catalan and Spanish complex predicate
structures the T-auxiliary does not trigger obligatory incorporation.
According to G&H, the fact that causatives are T-aux is supported by
the following ungrammatical examples, which follow from independent
ordering and distribution restrictions on auxiliaries - cf. 3.4.1 and 2.4 -
(G&H(25)and(26):
(214) * /e fais [ vp avoir [ yp travaille Marie II
T-aux N-aux
(215) *Imad&[yp Mary be wording I
T-aux T-aux
*
As they point out for English, if the causative were not considered an
auxiliary, but a lexical verb, it would have no access to the auxiliary
structure of its complement. The same holds for Catalan and Spanish:
(216) *ßä [ yp haver [yp treballat la Maria II
(217) *Hace { yphaber / yp trabajado (a)Maria II
As noted, the complement of a causative is a CTC, which is what
defines it, basically, as a T-aux. If the causative is not a structural case
assigner - such as make/let - so that the subject in the complement cannot
be assigned a theta-role, " the grammar contains alternative strategies to
ensure that the complement of the causative is a CTC" ( p. 56 ). As already
explained in previous sections - especially 3-4.1.1 -, this is basically the
reason why IP is still postulated in the syntax, so that the I node can
absorb Case F, and the AGR in INFL may function as a pronominal, taking
up the external theta-role assigned by the main verb. VP construal is
ensured in LF.
G&H account for the different types of causatives - cf. section 2.3, and
basically Kayne (1975) - in different ways. Basically, the /«r-construction
is seen as a doubling of the external argument, as in passive by- phrases;
the FI construction is regarded as involving Aux-to-Comp - V raises to
assign Case to its own subject -, because faire is not a structural case
assigner. The external theta-role is assigned to the dative NP embedded
under IP2 - (lôô)p. Ô9-.
(21Ô)
lire ej aux enfants
Quoting G&H: " Silice V1 T-marks IP2, there is no barrier between the
clitic tej and its trace. IP2 receives no case from faire and is construed as
VP in LF. V2 is then part of the same T-chain as the matrix causative. Since
V2 bears the same index as the matrix verb which carries the clitic, ej is
antecedent-governed." (p. 59)
G&H do not classify modals «as either N-auxiliaries or T-auxiliaries.
Nevertheless, given the definition of CTC; repeated here:
(219) An XP which is assigned a T-role is a Complete Thematic
Constituent (CTC): all theta-roles associated with X, the head
of XP, are assigned internal to XP.
Root modals can be claimed to be T-auxiliaries: when they function as
auxiliaries, despite the fact that they assign an external theta-role, this
same theta-role is borne by the PRO in their complement, which is a
possible interpretation of the CTC requirements. It is true that root modals
do not satisfy another of the characteristics of T-auxiliaries; i.e. that they
may be structural case assigners, but G&H do assume that they have a Case
F to assign which may be absorbed by INFL and give rise to a complex
predicate structure, as explained in the previous section.
The same cannot be claimed for epistemic modal verbs , which do
not impose selectional restrictions on the matrix subject -i.e. they do not
assign an external theta-role, - cf. 2.3 especially Picallo (19Ô5) -- In this
sense, they function more like complex verb sequences. Picallo (1990)
argues against their being considered raising verbs and allows them to be
generated in a different, inflectional, node. Assuming this might lead to a
postulation of a different structure for epistemic modals - as in Picallo
(19Ô5)-(1990) ; i.e. not involving a biclausal structure. Nevertheless, it is
possible to still grant them a "typical" complex predicate structure in the
syntax - i.e. a biclausal structure; cf. 2.2.1 - and assume that syntactic
incorporation takes place following the HMC; as there is no need for the
whole VP to move. Recall that Baker's main reason to claim that the whole
VP moves was the fact that the subject of a transitive causative complex
predicate surfaced as oblique. He nevertheless does not posit such a
movement for intransitive structures - cf. section 3.3.2 -. Note that in modal
complex predicate structures, precisely, subjects do not surface.
In favour of Picallo's hypothesis is the fact that one of the original
claims - cf. 2.1.2 - for postulating a biclausal status for modal sequences is
not obtained with modal epistemic verbs. Quoting Picallo (1990):
" Epistemic modals are the only type of predicates ( verbal or adjectival)
that do not assign a theta-role to the subject position, and cannot
subcategorize for [+AGR] sentential complements (indicative or
subjunctive)." (p.295):
(7ô)a. *P¿>¿ que no em trobin quan tornin 46
b. *ßeu que en Pavarotti canti a IFstadi Olimpio
(79) *Bebe que José María no se acuerde demi
Moreover, the fact that they do not contribute to theta-role
assignment of the external argument of the predicate seems to me a crucial
piece of evidence to classify them as constituting complex verbs as opposed
to complex predicates.
Given the above observations, I will claim that epistemic modals are
N-auxiliaries, displaying the typical behaviour of an N-aux in that they
allow for the percolation of the theta-role assigned by the main verb. Now,
if epistemic modal verbs are claimed to be neutral auxiliaries, then they
should trigger obligatory incorporation, as other N-auxiliaries in Catalan
and Spanish do. This prediction seems to be borne out by the fact that
dividing complex predicates made up of epistemic modals - or modals with
an epistemic modal reading - gives rise to ungrammatical structures,
whereas those made up of root modals do not:47,4ô
Consider some examples of interruption and preposing in Catalan and
Spanish with epistemic and root modal verbs:
a) (epistemic)
(222) *M'han dit que en Pavarotti M de cantar a liceu, però
cantar no deu
(2 2 3) *?Peu sempre marjztr desprès de classe
(224) * Encara que el Joan sigui mal conductor, tenir un accident per
la Meridiana no pot
(2255* Aunque Pepita sea mala cocinera, quedar tan mal con sus
invitados no puede
(226)?? Debe a menudo salir corriendo después de classe
b) (root)
(227) a.(?) No podria, de cap manera, viiwe a Anglaterra
X>.ll Bisabet.pot dissenyar plans d édifias amb el seu ordinador
pero imprimir-los no pot.
(2 2 ô)a.? Tengo unas ganas de poder un dia dormir 12ñoras seguidas/
b. ? Juan Alberto puede pilotar aviones con su ordenador, pero
los de verdad, pilotarlos no puede
I repeat here the examples where VP anaphora was considered a
property of modals being auxiliaries. Note that these are also root modals:
( 196) En Pere voldria afinar el piano
a. però no pot/volIyp fer-ho I
b. però nopotsvol
c. *però no creu/deddeiK
( 197) Pedro querría afinar el piano
a. pero no quiere/puede / ¡/placerlo /
b. pero no quiere/puede
c. * pero no cree /decide
A note on English T- and N- auxiliaries
As postulated also in G&H's framework the classification into T-
auxiliaries and neutral auxiliaries is also applied to English. The patent
differences with Catalan and Spanish obviously call for an explanation. G&H
themselves point out that " auxiliary verbs do not have the same syntactic
properties over or within languages" (p.47). It was already noted - cf.
section 2.4 - that Italian essere, although it is a T-auxiliary, may refrain
from assigning its T-role, while another T-auxiliary, English ¿>e, may not. It
is evident that the fact that there are properties which are shared and
constant over languages is what has brought G&H to their dual classification
of auxiliaries.
In this section I have linked the notion of W-auxiliary with the
triggering of obligatory incorporation; i.e. in complex verb structures - and,
possibly in epistemic verb structures -. English N-auxiliaries obviously do
not trigger incorporation - cf. the differences pointed out in section 3.3-1.4-.
English HAVE, the paradigmatic N- auxiliary, when occurring in equivalent
complex verb structures, does not prevent interruption nor disallow
preposing; moreover, there is not even marginality in the preposing
examples of complex verb sequences in English, so incorporation is utterly
unmotivated :
(2 2 9) My son said he had passed his examination and
passed it he has
(2 30) *E! seu fill va dir-Jj que havia passat l'examen i passat 1 ha
(231) * Su hijo le di/o que haMa pasado el examen y pasado lo ha
Note also that English has the support vert» do which has no
equivalent in Catalan or Spanish, since even in preposing examples of
simple past - which needs do-support in English -, the use oí a substitute
verb in Catalan and Spanish does not give rise to grammatical results:
(232) I/o said ño would jump ovor tno balcony., andjump no did/
(233) * Va dir quo saltaria pol balcó i saltar va (fer)
(2 34) *I)jjo quo saltar/à por ol balcon y saltar t'ñisoj
The fact that the past participle and the auxiliary form an X-0
constituent - i.e. as well as in the other Catalan complex verb sequence
analyzed in this thesis, anar + infinitive - is a language particular fact
about Catalan and Spanish. Note, though, that this cannot be extended to
other Romance languages, nor to pro-drop languages: as for the first
observation, French, a Romance language, is obviously a non-V-
incorporation language:
(235) p£ 'aipas /jamais/ encoró vu uno cnoso aussi atroco
And as for the second observation, Italian, a Romance and pro-drop
language, is also a non -obligatory -V -incorporation language - cf. also
section 3.3.I.4 for a summary of Belletti (1990) -:
(236) Non no mal visto queseo
The lack of lexical meaning of the verbal elements ñavor/Mbor has
already been alluded to in claiming their status as a clitic - cf. section
3.3.I.I -, and as a factor which, to my mind, is crucial in the triggering of
obligatory incorporation. This stands in clear constrast with English, French
and Italian where the equivalents for HAVE have not lost their lexical
meaning.
As regards complex predicates in English an important difference
must be noted. Firstly, modals are INFL elments, so no equivalent modal
"restructuring" constructions are found in the language. Aspectual complex
verb sequences give the same results in English, Catalan, and Spanish,
which is predicted by the fact that there is no incorporation in any of the
three languages in these sequences.49,50
(237) / said I would begin to dance,, and to dance 1 will begin
(2 30) Us ne dit çue començaria a bailari a bailar començaré
(239) Os dije çue empezaría a bailar y a bailar empezaré
As regards causative sequences in English, it has already been
explained and shown - cf. 2.3.1 - that there is no equivalent configuration
to the Romance causative complex predicate sequences. Nevertheless, G&H
posit that English causatives are T-auxiliaries which have the ability to
assign structural case to the subject of their SC complement.
Again the hypothesis of non-incorporation accounts for the fact that
preposing gives grammatical results in English:
(240) We tñougnt sne would ma^e mm sing, and sing sne made Mm
But interruption results in ungrammmaticality:
(241)* Sne mal-es always ner nusband sing
This may be accounted for by StowelK 19Ô D's Case adjacency Priciple
which holds for English, and would rule the sequence out.
Note that Spanish and Catalan perception verbs - also noted as
candidates for complex predicate sequences - may also be assumed to be
structural case assigners, which is not incompatible with their possible T-
auxiliary status:
(242) Juan vio a María salir del coche
(243) En Joan va pleure (à) Ja Maria sortir del cotxe
Note that non-syntactic incorporation predicts interruption and
preposing to be ( at least marginally) acceptable:
(2 44)a.? Juan creyó ver a María salir del coche, pero en realidad
salir no la vio
b. Juan vio sin ninguna duda a Maria salir del coche
(245) a. ? En Joan va creure que vela a la Maria sortir del cotse.però
de fet sortir no la va veure
b. En Joan va veure sense cap mena de duMe a la Maria sortir
delcotse
Notes to Chapter 3
(1) This brief "note" does not intend to go further than this observation. See
section 3.3.1.1 for comments and reference to ctitJazatÀw, which is,
obviously "behind" this observation.
(2) As will become clear in the la ter sections, he does not use the term
JnœrporatÂw in Baker's sense.
(3) Aarts ( 19Ô9) argues against analyses of this type for such constructions
- and others - positing instead a rightward NP movement adjunction to VP
- instead of incorporation - and small clause subcategorization.
(4) Crucially, as will be noted in section 3.3.1, this principle does not hold of
the verbal sequences under analysis in that section.
(5) This, again, will not be relevant for the sequences under analysis for
Romance languages - cf. section 3.3.1 -, where morphology is not as
productive as in agglutinating languages.
(6) Baker's account of passive is summarized in Note (3D. There he
proposes abstract (LF) - mœrpûratÀw.
(7) See Spencer (forthcoming), and Ouhalla (1990) for a review and critical
analysis of Baker ( 1986).
(Ô) His theory is based on the barriers framework - cf. section 1.2.2 -, with
several modifications that will be included if they are relevant to the
explanation.
(9) See section 3.3.1.4 for some comments on language variation; namely,
differences among Italian, French, and English with respect to Catalan and
Spanish.
( iô) Baker uses S/'S: notation, I wiii keep to this usage for simplicity
reasons.
( 11) See (34) in the text for the notion of distinctness.
( 12) In 3.4 we will be led to the assumption that in complex verb
structures, T-marking- G&H ( 19ÔÔ)- is a sufficient requirement for
incorporation in complex verb sequences; there will be no need to allude to
theta-indexing of VPs.
( 13) I must note that I will be omitting for the most part a more complex
phrase structure posited basically in Pollock ( 19Ô7), until Chapter 4, for
simplicity reasons. The structure I am referring to is one which includes
more functional nodes in the phrase structure. See note (25) below.
(14) Note that the VA sequence is not found in Spanish; the "simple past" is
expressed only through the use of the inflected form of the main verb:
(i) Imprimió el documento
(ii)a. Imprimí el document
b. Va imprimir el document
The use of the verb ir with an infinitive in Spanish, Va a imprimir, is a
future aspectual periphrastic sequence, or a sequence of two main verbs -
cf. section 2.1.1, Test 7-
(15) This could obviously be argued to be a counterexample to the modified
HOC ,(54) in the text.
( 16) As already noted in the review of literature on complex predicates,
there are many views on how clitic+host sequences are to be characterized;
the two basic positions are represented by Kayne (1975), Rizzi (19Ô2),
Aoun ( 19Ô5) - movement - among others, and Borer (1964), Jaeggli ( 19Ô2)
- base-generation - among others.
(17) In this section - cf also note (13) - I will not make use of a more
complex structure where participles are dominated by a non-lexical node,
as in Belletti (1990) - although I will mention her structure in section
3-3.1.4 - The same issue arises in 3.4, as will be noted, cf. Note (25) -
( 16) In the revised version of the paper - Manzini ( 19ÔÔ) - an identical
structure is not posited. Nevertheless, I have included it because it is in line
with the structure proposed in this thesis for complex verbs in Catalan and
Spanish, as well as English.
(19) This obviously does not apply to VP-adjunction in extraction
structures, a basic mechanism in Barriers - cf. section 1.2.2 -
(20) I will not go into the details of the type of indexing involved in these
VP complements; nevertheless, as stated in 3-4, T-marking will be assumed
to subsume theta-indexing in complex verb structures
(21) pas- may be argued to be a clitic in these constructions - i.e. where it
"interrupts" a complex verb -, as it clearly does not bear stress.
Nevertheless, ií pas follows the complex verb sequence, it may bear stress
-: None vist pas la nena.
It must also be noted that there is another element which may
marginally interrupt the sequence both in Catalan and in Spanish, ni:
(i)? No nan ni provat la sopa
(ii)? No nan ni probado la sopa
As pointed out to me by J.M. Brucart, both of these elements are
negative scope-markers, a fact which may bear on the explanation of their
position next to a lexical head; i.e. a head which may be negated.
(22) A possibly related fact may come from the English infinitival form,
where to * V are not morphophonologically amalgamated, and a certain
degree of non-cohesiveness seems possible:
(i)?? / want tú really go there/
Exampies in Pollock (19Ô7), to which he refers for other reasons may
bear on this issue - (27) in Pollock - :
(ii) To hardly understand Italian after years of hard work moans you
have no gift for languages
(23) in the case where a verb is not subcategorized for, or does not select,
a VP, LF incorporation is not necessary - possibly not even needed, as will
be argued for in 3.4 -.
*
(24) This structure is based on recent proposals summarized in Chapter 4 -
especially section 4.2 -. Basically, the INFL (I) node - cf. 1.1, 1.2.2 - which
was assumed to be the head of S ( = IP ) is now fleshed out into two other
distinct functional heads, AGR(eement) and T(ense), each of which has its
own maximal projection, AGRP, and TP. The order of these categories is a
matter of debate. I have included this specific structure in an attempt to
answer the problem of the bearing of inflectional affixes by V1 and not by
V2 in complex verb sequences. For further explanation and comments on
this new structure see Chapter 4, section 4.2.
(25) Belletti's structure includes a functional node, AGR, above the VP
participle projection. I have included the whole structure for the sake of
faithfulness to the text which is being commented on. Nevertheless, I will
be disregarding the possible intervention of a functional node between the
auxiliary and the non-finite form in a complex verb sequence until Chapter
4 , where it is briefly touched upon in section 4.3. As argued in section
3-3.1.2,1 am assuming a consecutive VP-VP structure. - cf. also Notes (13)
and (17)-.
(26) By referring to "content" lexical status I am deliberately not alluding to
the difference between lexical/functional nodes. Note that although not
"content" lexical, I assume that auxiliaries in complex verb sequences are
generated in V lexical nodes.
(27) Note that the ? marks are included in parentheses because there is
variation in the acceptability of all of these examples; for some speakers
they are wholly acceptable, for others they are not.
(20) Note that this differs from Chomsky ( 19ô6b)'s analysis of passive: BE
moves to INFL, but the lexical verb remains in its original position - cf.
section 1.2.2 -.
(29) Baker distinguishes between Chichewa-A, as opposed to Chichewa-B,
which differ in their case parameters. Only Chichewa-A follows rule (232)
in text.
(30) Chichewa and Malayalam have another crucial objecthood test: verb
object agreement. They also have passivization, but they lack cliticization.
Actually, clitics may be regarded as a kind of object agreement.
(3D As Baker points out - p. 463, in. 36" -, this second test does not work
for all Romance languages - cf. also section 3.4 -.
(32) Bakers analysis of passivizatJon, a crucial GF changing phenomenon
also involves the assumption that there is abstract Jncûrporatjon, in this
case of V-to-I; i.e. of a lexical to a non-lexical node, both in English and
Romance. Passivization is, thus, argued to be a syntactic and not a
morphological phenomenon - Note that here I am not alluding to the
distinction made by Wasow (1977) between lexical/syntactic passive, as
noted in section 2.2.1, and cf. also 3.1 and the explanation of the Mirror
Principle in that section -. The fact that both morphological and copular
passive are attested in the languages of the world is regarded as - on a par
with causativization - as a consequence of the fact that incorporation may
apply at different levels of the grammar: in Chichewa, for instance,
passivization is morphological and thus an instance of incorporation from D-
structure to S-structure; but in English or Romance, passivization is copular
so an instance of LF incoporation. Both, nevertheless - as required by the
UTAH - have an identical D-structure.
Baker assumes that the PASS morpheme is a nominal element,
generated in INFL, as in (i), and is assigned - i.e. it does not absorb, as in
other accounts - the external theta-role.
Vi NP
steal
The fact that a thematic nominal element is present is argued by
Baker on the basis oí evidence from Binding, Control and secondary
predication; i.e. it is syntactically active. Note that the fact that the external
theta-role is assigned to the PASS morpheme in INFL frees the NPsubj
position from thematic status and allows NP-movement.
The way a nominal element in INFL acquires case varies. In languages
where this nominal element is I it is by the asssignment of case by the V
which incorporates into INFL. The implicitly assumed principle that Baker
makes explicit: No category may assign case to itself implies that if the
PASS morpheme is I, it will not get case unless the verb has a case to assign
- which predicts the fact that only accusative-assigning verbs may be
passivized. Baker accounts for the fact that there are languages which allow
passivization of non-accusative verbs by positing variation in the category
type of nominal element in INFL.
(33) No reference to the structures showing the application of incorporation
will be given until section 3-4.1.2.
(34) See Pollock ( 19Ô7), and Chapter 4 for V-rnovement in French and
English.
374
(35) Note that the actual T-chain formation has been illustrated above to be
equivalent in English, French, Catalan and Spanish, previous to
incorporation.
(36) G&H give a parallel explanation for complex predicate structures
where the first element is considered an auxiliary - cf. section 3.4.2.2 -.
(37) As already noted in 3-3.1.4, when summarizing Belletti (1990), for
some past participle sequences there is assumed to be a functional node
dominating the past participle. This is explicitly argued for in several recent
works such as Kayne (19Ô7), Pollock. (19Ô7), G&H (19ÔÔ), Chomsky (19ÔÔ),
Drijkoningen (19Ô9). The main difference fort most of these authors lies in
the fact whether there is participle agreement or not; i.e. in the former case,
there is a functional node, in the latter, there is not. In Spanish G&H
explicitly allow VP in the syntax because there is no agreement, as shown
in their examples - Í102) p. 69 -, as opposed to French:
(i) a. p JesêJprises
b. Losfj& tomaäo
Note that Catalan allows agreement:
(ii) Le hepres/pres&s
For further (brief) consideration of this matter see Chapter 4, especially
section 4.3-
(30) This holds as long as we assume that there are no functional nodes
intervening between the two verbs in the sequence. See Note (36).
(39) Zagona (19ÔÔ) notes that. VP-preposing is a type of A-bar movement,
and she assumes it may move to Spec- CP. Nevertheless, she herself gives
an alternative of Topicalization and and a null operator as in (i) - (p. 125) - ,
which would account for the Subjacency effects that she notes for English
preposing structures :
(i)... and [TOPIC IVP leave] Icp Oi lip he will ei ] 3 ] )
(40) Another passive, the impersonal passive is possible, and used: (i) Els
llibres es van fer llegir ; (ii) Los libros se hicieron leer.
(41) As explained in 2.4, and further considered in 3.4.3, if Catalan and
Spanish BE are here considered T-auxiliaries, independent constraints on
auxiliaries explain the ungrammatical examples in Spanish and Catalan;i.e
basically the fact that no two auxiliaries with the same function may
appear in a sequence.
(42) G&H- footnote ô, pg. 4ô- make a difference between simple and
complex tense as in (i) and (ii): avoir may denote either a simple or
complex tense, whereas A?P<? may only denote a complex tense, as in (ii)*
(i) J'ai vu Pierre à quatre heures/souvent
(ii) I have seen Peter *at four o'clock/ often
(43) This has been explained and illustrated - with clitic climbing and
theta-role percolation examples - in 2.4; the main idea behind it being the
fact that it links up tense as an operator on a theta-domain.
(44) This use is also allowed in Catalan under restricted circumstances.
(45) An alternative explanation as regards modals - which I will not pursue
in this thesis - is in line with Picallo (1985)-(1990);i.e. granting root modals
an adjunct predicate status, an idea that springs from Zubizarreta (1952),
and which Pollock (1967) - as Picallo (1990) notes in a footnote - also
illustrated with the following examples:
(i) Pierre a voulu partir
(ii) Pierre est parti volontairement
Modal verbs seem to contribute to the interpretation of the predicate in an
adverbial fashion rather than by being direct theta-role assigners
(46) In Spanish the equivalent is possible: (i) Puede que no me encuentren
cuando vuelvan. Picallo( 1990) refers strictly to Catalan. The hypothesis in
this section may also only apply to Catalan, although in respects to be seen
below, epistemics in Catalan and Spanish seem to function identically
(47) Note first that in those examples given in 33.2.1, the sentences where
interruption of complex predicates is acceptable, the first element is a root
modal, or has a root, interpretation. The fact that Spanish deber, as opposed
to Catalan deure, may also have a root reading explains example ( ) in
3.3-2.1 from H&RU9Ô4))-
(40) Note that there is an important exception, the negative particle may
interrupt, and gives marginal results, even if the reading is epistemic, both
in Catalan and Spanish. :
(i)? En Guillem deu no saber què fer
(ii)? Guillermo debe no saber qué hacer
As will be noted in Chapter 4, in the present, phrase structure proposals à-
la-Pollock, the negative particle status has a special status. I will briefly
point out some possibilities in the chapter Prospects
(49) The fact that some speakers of Catalan and Spanish consider the
examples not wholly acceptable obviously calls for an explanation.
(50) The fact that the preposing construction for Catalan and Spanish is not:
"... ballar començaré a" / "... AyJar empezaré a" may be linked to the lack
of preposition stranding in the two languages.